TECHNOFREAKS

2007-10-07T00:33:00.000-07:00

Robot Demos:
Interactive control of the DLR hand
Human control of robotic systems with a large number of DOFs is often difficult due to the overwhelming number of variables that need to be specified. We developed a method for controlling such systems using only a small number of input variables. Motivated by prosthetic applications, we applied our method on the control of a robot hand. Come and control the DLR robot hand using the mouse or a wiimote!

An insight into R,LAB's latest robotics platform, the SmURV
Robots either cost a lot of money or require hundreds of hours to be built; sometimes even both. The SmURV is a comparatively cheap basic robotics platform that was developed at R,LAB and is made up of off-the-shelf computer components. Come and see how easy it is to build one yourself!

Tactical Teams
We present a robotic system that integrates person and gesture recognition with speech recognition and synthesis. This results in a natural, hands-free interactive system.

Robotic Video Games
A video game using Nintendo Wii Remotes as controllers for Sony AIBO dogs playing soccer. This is an interactive demo, so viewers will be allowed and encouraged to pick up a Wii Remote and join the fun!

Robot learning from Demonstration
Programming a robot to perform a task can be a long, arduous undertaking. We are exploring ways in which users can demonstrate tasks to robots in an intuitive manner and the robot learns to perform the task. Participants can interact with the robots and attempt to teach them new behaviors.

Graphics Demos:
ChemPad
Molecules are inherently three-dimensional objects that are represented by chemists on paper and classroom blackboards by a system of two-dimensional notations. ChemPad, a new Tablet PC application with a pedagogical focus, generates 3D molecular structures from hand drawn digital ink rather than from traditional molecule construction interfaces.
ChemPad allows chemists to sketch molecules in a quick and natural fashion to generate 3D models. This capability gives ChemPad pedagogical value as shown by our user study of a hundred organic chemistry students who had difficulty with 3D chemistry thinking and used ChemPad to overcome that hurdle.

Math Recognition & Error Correction
The Math Error Correction project is concerned with the user interface aspects of entering mathematical notations on a pen-based computer. The sometimes ambiguous nature of mathematical notation and the irregular, imprecise and ambiguous nature of human handwriting conspire to produce inconsistencies between what the human and the computer perceive. We have developed a sophisticated mathematic handwriting recognizer, and a handful of visualization techniques to aid the user in identifying parsing errors. Our demo also includes techniques for correcting the errors and for incorporating the math into a more general writing environment.

MathPad
MathPad2 is a prototype Tablet PC application that allows users to make dynamic illustrations by combining handwritten mathematical expressions with sketched drawings. This combination allows for animated illustrations that aid users in understanding mathematics, physics and engineering concepts. Users can also graph and solve equations as well as simplify and factor expressions with a simple gestural interface. The prototype makes use of Mas a computational back end and also has a sophisticated, trainable math recognition front endUsers can label their sketched diagrams and the system uses these labels to determine the associated mathematics. Output from Matlabis used to drive simple animations so users can observe diagram behavior.
Unlike other mathematics systems that require typed linear notations, MathPad2 uses natural 2D mathematical notation. In addition, the experience islive and interactive in contrast to a static paper notebook. With MathPad2, we have a powerful electronic problem solving notebook that utilizes pencil-and-paper style input with pen-based computing that can dramatically change how students learn and work.

Diagrammer
Our goal is to make the process of producing 2D diagrams much easier and more efficient. The observation that most diagrams can be sketched with a pen in a few seconds, while other interfaces (e.g., keyboard and mouse) wouldrequire much more time and seem tedious by comparison is why we believe a pen-based interface is probably the best interface for producing the kinds of diagrams most people need. However, an initial sketch can be many steps away from a completed diagram and what is needed is an end-to-end process that preserves the efficiency and ease with which an initial sketch can be made.

2007-09-20T06:32:00.000-07:00

Skeletonization (Roadmap) Methods

Skeletonization methods collapse the configuration space into a one-dimensional subset, or skeleton. They then require that paths lie along the skeleton. If the initial and goal points do not lie on the skeleton, short connecting paths are added to join them to the nearest points on the skeleton.

To be complete for motion planning, skeletonization methods must satisfy two properties:

1. If S is a skeleton of free space F, then S should have a single connected piece within each connected region of F.

2. For any point p in F, it should be "easy" to compute a path from p to the skeleton.

Roadmaps are a global approach to motion planning. It consists of modeling the connectivity of the robot's free space as a network of one-dimensional curves, called the roadmap, which lies in the free space or its closure. (NOTE that weird things happen at the boundaries depending on which option you choose.)

Once a roadmap R has been generated, it is used as a set of standardized paths. Path planning is then reduced to connecting the initial and goal configurations to points in R, and searching R for a path to connect those points.

The kinds of skeletonizations are visibility graphs, Voronoi diagrams and roadmaps, which consist of silhouette curves and linking curves. (This formulation due to Russell and Norvig; note that Latombe calls all of these things "roadmap methods.")

Visibility Graphs

The original roadmap method, visibility graphs apply to 2D configuration spaces with polygonal C-obstacle regions. The graph is nondirected and the nodes are the initial and goal configurations plus all the vertices of the C-obstacles. The edges are all the straight line segments connecting two nodes that do not intersect the interior of the C-obstacle region (all lines that don't go through obstacles). The graph can be searched for the shortest semi-free path; this path will always be polygonal. Visibility graphs are complete.

Voronoi diagram

A roadmap method based on retraction. A continuous function of C_free is defined onto a one-dimensional subset of itself. The Voronoi diagram consists of those curves in the space that are equidistant from two or more obstacles; these curves form the skeleton.

Put poorly in another way, he Voronoi diagram is the set of all free configurations whose minimal distance to the C-obstacle region is achieved with at least two points in the boundary of the region. This method yields free paths which tend to maximize the clearance between the robot and the obstacles.

When the C-obstacles are polygons, the Voronoi diagram consists of straight and parabolic segments. The initial and goal configurations are mapped to q_init and q_goal by drawing the line along which the distance to the boundary of the C-obstacles increases the fastest.

Cell Decomposition

A global approach to motion planning. The intuitive idea is to break the space down into a finite number of discrete chunks.

Cell decomposition breaks free space down into simple regions called cells such that any path between any two configurations in a cell can easily be generated. A non-directed graph representing the adjacency relations between cells is then constructed and searched. This is called the connectivity graph or adjacency graph. The outcome of the search is a sequence of cells called a channel. There are two cell decomposition methods:

· approximate cell decomposition -- Cells have predefined shapes (such as triangles or trapezoids) whose union is strictly included in the free space. The boundary of a cell does not characterize a discontinuity of some sort and has no physical meaning. The quad-tree method recursively decomposes a rectangular cell into four smaller cells until the cells lie entirely in the free space or reach some resolution. These methods are sound but not incomplete.

· exact cell decomposition -- The free space is decomposed into cells whose union is exactly the free space; this is a complete method. Cells in this form of decomposition are called cylinders.

Potential Fields

A local approach to motion planning. The goal configuration generates an attractive potential that pulls the robot toward it; C-obstacles generate repulsive potential that pushes the robot away. The negated gradient of the total potential is treated as an artificial force applied to the robot, considered the most promising direction.

Potential fields are very efficient but suffer from local minima. Two approaches overcome this:

· design potential functions with no local minima other than the goal configuration

· complement the basic potential field approach with mechanisms to escape from local minima

Landmark-Based Navigation

Landmark-based navigation assumes that the environment contains easily recognizable, unique landmarks. A landmark is modeled as a point surrounded by a field of influence. Within this field of influence, the robot knows its position exactly. Outside of all fields of influence, the robot has no direct position information.

Landmarks can be used to deal with uncertainty in motion so, for example, if a robot knows it is apt to move with slight uncertainty as it travels, it can aim for a central point in a field of influence, knowing that the actual cone it maps out will intersect with that field no matter how off its projections it is.

Configuration Space

For a robot with k degrees of freedom, the state or configuration of the robot can be described by k real values. These values can be considered as a point p in a k-dimensional configuration space of the robot.

Configuration space can be used to determine if there is a path by which a robot can move from one place to another. Real obstacles in the world are mapped to configuration space obstacles, and the remaining free space is all of configuration space except the part occupied by those obstacles.

Having made this mapping, suppose there are two points in configuration space. The robot can move between the two corresponding real world points exactly when there is a continuous path between them that lies entirely in configuration free space.

Generalized Configuration Space

The term generalized configuration space is used to describe systems in which other objects are included as part of the configuration. These may be movable, and their shapes may vary.

There are several ways of dealing with robot planning when there are several moving or movable objects. These are:

1. Partition the generalized configuration space into finitely many states. The planning problem then becomes a logical one, like the blocks world. No general method for partitioning space has yet been found.

2. Plan object motions first, and then motions for the robot.

3. Restrict object motions to simplify planning.

Extensions of the Basic Problem

Multiple Moving Objects

Obstacles may be moving. There may be multiple robots. The robots themselves may be articulated (i.e., made of rigid objects connected by joints).

The first extension, and possibly the second, requires time to be explicitly considered, which gives rise to configuration-time space in which a path must never go back along the time axis, and some constraints on the slope and curvature of the path are given due to constraints on velocity and acceleration.

The second and third extensions yield configuration spaces of arbitrarily large dimensions. Planning can be done in a composite configuration space which is the cross-product of the individual configuration spaces (this is called centralized planning), or another method called decoupled planning can be used to plan the motions more or less independently and interactions are only considered in the second phase of planning.

The decoupling method fails when interaction is critical -- in the example of two robots trying to switch places in a narrow corridor, for example. When dealing with articulated robots, the dimension of the configuration space grows as the number of joints.

Kinematic Constraints

A robot's motions may be restricted by kinematic constraints. There are two kinds:

holonomic constraints

A holonomic equality constraint is an equality relation among the parameters of the minimally-represented configuration space that can be solved for one of the parameters. Such a relation reduces the dimension of the actual configuration space of the robot by one. A set of k holonomic constraints reduces it by k. For example, a robot limited to rotating around a fixed axis has a configuration space of dimension 4 instead of 6 (since revolute joints impose 2 holonomic constraints).

nonholonomic constraints

A nonholonomic equality constraint is a non-integrable equation involving the configuration parameters and their derivatives (velocity parameters). Such a constraint does not reduce the dimension of the configuration space, but instead reduces the dimension of the space of possible differential motions.

For example, a car-like robot has 3 dimensions: two for translation and one for rotation. However, the velocity of R is required to point along the main axis of A. (This is written as - sin T dx + cos T dy = 0.)

The instantaneous motion of the car is determined by two parameters: the linear velocity along the main axis, and the steering angle. However, when the steering angle is non-zero, the robot changes orientation, and its linear velocity with it, allowing the robot's configuration to span a three-dimensional space. Restricting the steering angle to pi/2 restricts the set of possible differential motions without changing its dimension.

Nonholonomic constraints restrict the geometry of the feasible free paths between two configurations. They are much harder to deal with in a planner than holonomic constraints.

In general, a robot with nonholonomic constraints has fewer controllable degrees of freedom than it has actual degrees of freedom; these are equal in a holonomic robot.

Uncertainty

A robot may have little or no prior knowledge about its workspace. The more incomplete the knowledge, the less important the role of planning. A more typical situation is when there are errors in robot control and in the initial models, but these errors are contained within bounded regions.

Grasp Planning

Many typical robot operations require the robot to grasp an object. The rest of the operation is strongly influenced by choices made during grasping.

Grasping requires positioning the gripper on the object, which requires generating a path to this position. The grasp position must be accessible, stable, and robust enough to resist some external force. Sometimes a satisfactory position can only be reached by grasping an object, putting it down, and re-grasping it. The grasp planner must choose configurations so that the grasped objects are stable in the gripper, and it should also choose operations that reduce or at least do not increase the level of uncertainty in the configuration of the object.

The object to be grasped is the target object. The gripping surfaces are the surfaces on the robot used for grasping.

There are three principal considerations in gripping an object. They are:

· safety -- the robot must be safe in the initial and final configurations

· reachability -- the robot must be able to reach the initial grasping configuration and, with the object in hand, reach the final configuration

· stability -- the grasp should be stable in the presence of forces exerted on the grasped object during transfer and parts-mating motions

Example: peg placement. By tilting a peg the robot can increase the likelihood that the initial approach conditions will have the peg part way in the hole. Other solutions are chamfers (a widening hole, producing the same effect as tilting), search along the edge, and biased search (introduce bias so that search can be done in at most one motion and not two, if the error direction is not known).

Computational Complexity

These theorems are taken from Latombe's book (see Sources).

Theorem 1 (Polyhedral bodies) Planning a free path for a robot made of an arbitrary number of polyhedral bodies connected together at some joint vertices, among a finite set of polyhedral obstacles, between any two given configurations is a PSPACE-hard problem.

Put another way, planning a free path in a configuration space of arbitrary dimension among fixed obstacles is PSPACE-hard.

Theorem 2 (Joints) In the absence of obstacles, deciding whether a planar linkage in some initial configuration can be moved so that a certain joint reaches a given point in the plane is PSPACE-hard.

Theorem 3 (Rectangular robots) Planning a path in the configuration space of a multi-bodied robot consisting of all rectangles is PSPACE-hard.

Theorem 4 (Planar arm) Consider a planar arm consisting of arbitrarily many links serially connected by revolute joints such that all the links are constrained to move in the plane. Planning a free path for such an arm among a finite number of polygonal obstacles, between any two given configurations is PSPACE-hard.

Theorem 5 (Upper bound for fixed dimension) A free path in a configuration space of any fixed dimension m, when the free space is a set defined by n polynomial constraints of maximal degree d, can be computed by an algorithm whose time complexity is exponential in m and polynomial in both n and d.

This theorem is based on reducing the planning problem to the problem of deciding the satisfiability of sentences in the first-order theory of the reals. The important observation is that the complexity of path planning increases exponentially with the dimension of the configuration space.

Theorem 6 (Velocity-bounding) Planning the motion of a rigid object translating without rotation in three dimensions among arbitrarily many moving obstacles that may both translate and rotate is a PSPACE-hard problem if the velocity modulus of the object is bounded, and an NP-hard problem otherwise.

Theorem 7 (Uncertainty) Planning compliant motions for a point in the presence of uncertainty, in a three-dimensional polyhedral configuration space with an arbitrarily large number of faces, is an NEXPTIME-hard problem.

Reducing Complexity

It is possible to approximate the real problem before giving it to the motion planner; it is reasonable to trade generality for improved time performance.

One way of reducing complexity is by reducing the dimension of the configuration space. This can be done by replacing the robot by the surface or volume it sweeps out when it moves along r independent axes. This corresponds to projecting the m-dimensional configuration space along r of its dimensions. The projected configuration space has only m - r dimensions. (As an example, imagine a robot with an extended arm as a disk; then you don't need to worry about rotations.)

Simplification heuristics make only local plans, by breaking the problem into subproblems. For example, many localities are stereotyped situations, such as moving through a door or turning in a corridor.

Robot Architectures

Brooks' Subsumption Architecture

This approach breaks the problem down into task-achieving behaviors (such as wandering, avoiding obstacles, or making maps) rather than decomposing it functionally (into sensing, planning, acting).

In subsumption architectures, levels of competence are stacked one on top of another, ranging from the lowest level (object avoidance) to higher levels for planning and map-making. Higher levels may interfere with lower levels, but each level's architecture is built, tested and completed before the next level is added. In this way the system is robust and incrementally more powerful.

Individual levels consist of augmented finite state machines connected by message-passing wires. Higher levels may inhibit signals on these wires, or replace them with their own signals; this is how they exercise control over more basic functions.

(For more information, see the AI Qual Summary on Agent Architectures.)

Shakey

Shakey was the forerunner of many intelligent robot projects. The Shakey system was controlled by PLANEX, which accepted goals from the user, called a STRIPS subsystem to generate plans, and then executed them via intermediate-level actions. These ILA's were translated into complex routines of low-level actions that had some error detection and correction capabilities.

After each action was executed, PLANES would execute the shortest plan subsequence that led to a goal and whose preconditions were satisfied. In this way, actions that failed would be retried and serendipity would lead to reduced effort. If no subsequence applied, PLANEX called STRIPS to make a new plan.

Situated Automata

An approach pioneered by Stan Rosenschein that begins with an explicit representation and reasoning system, but compiles it into a finite-state machine whose inputs come from the environment and whose outputs connect to effectors.

This compilation approach distinguishes between the use of explicit knowledge representation by the designers (the "grand strategy") and the use of explicit knowledge within the agent architecture (the "grand tactic"). Rosenschein's compiler generates FSMs that can be proved to correspond to logical propositions about the environment, provided the compiler has the correct initial state and physics.

The FLAKEY system at SRI used situated automata to navigate, run errands, and ask questions, and had no explicit representation.

2007-08-26T07:39:00.000-07:00

In this post we dwell on what are the requisite properties of robot,and the approach to program them accordingly.Also we ponder about some loop holes faced commonly,and about how to avoid them.

In general, robots should have the following qualities:

· High reliability -- if a robot fails, it should be able to recover or to call for help

· High speed -- a robot should perform its functions as quickly as needed

· Programmability -- the robot should be flexible and easily adaptable to various tasks

· Low cost

The ultimate goal is to build autonomous robots that accept commands telling them what to do, without needing to specify exactly how.

Task Planning

By virtue of their versatility, robots can be difficult to program, especially for tasks requiring complex motions involving sensory feedback. In order to simplify programming, task-level languages exist that specify actions in terms of their effects on objects.

Example: pin A programmer should be able to specify that the robot should put a pin in a hole, without telling it what sequence of operators to use, or having to think about its sensory or motor operators.

Task planning is divided into three phases: modeling, task specification, and manipulator program synthesis.

There are three approaches to specifying the model state:

1. Using a CAD system to draw the positions of the objects in the desired configuration.

2. Using the robot itself to specify its configurations and to locate the object features.

3. Using symbolic spatial relationships between object features (such as (face1 against face2). This is the most common method, but must be converted into numerical form to be used.

One problem is that these configurations may overconstrain the state. Symmetry is an example; it does not matter what the orientation of a peg in a hole is. The final state may also not completely specify the operation; for example, it may not say how hard to tighten a bolt.

The three basic kinds of motions are free motion, guarded motion, and compliant motion.

An important part of robot program synthesis should be the inclusion of sensor tests for error detection.

Motion Planning

The fundamental problem in robotics is deciding what motions the robot should perform in order to achieve a goal arrangement of physical objects. This turns out to be an extremely hard problem.

Motion Planning Definitions

Basic motion planning problem

Let A be a single rigid object (the robot) moving in a Euclidean space W, called the workspace, represented as Rⁿ (where n = 2 or 3). Let B₁, ..., B_q be fixed rigid objects distributed in W. These are called obstacles.

Assume that the geometry of A, and the geometries and locations of the B_i's are accurately known. Assume also that no kinematic constraints limit the motions of A (so that A is a free-flying object).

Given an initial position and orientation and a goal position and orientation of A in W, the problem is to generate a path t specifying a continuous sequence of positions and orientations of A avoiding contact with the B_i's.

(Basically, given a robot, a bunch of objects, a start state and a goal state, find a path for the robot to reach the goal state.)

configuration of object A

A specification of the position of every point in the object, relative to a fixed frame of reference. To specify the configuration of a rigid object A, it is enough to specify the position and orientation of the frame F_A with respect to F_W. The subset of W occupied by A at configuration q is denoted by A(q).

configuration space of object A

The space C of all configurations of A. The idea is to represent the robot as a point and thus reduce the motion planning problem to planning for a point.

dimension of C

The dimension of a configuration space is the number of independent parameters required to represent it as R^m. This is 3 for 2-D, and 6 for 3-D.

chart

A representation of a local portion of the configuration space. C can be decomposed into a finite union of slightly overlapping patches called charts, each represented as a copy of R^m.

Distance between configurations

The distance between configurations q and q' should decrease and tend to zero when the regions A(q) and A(q') get closer and tend to coincide.

Path

A path from a configuration q_init to configuration q_goal is a continuous map t : [0,1] -> C with t(0) = q_init and t(1) = q_goal.

free-flying object

An object for which, in the absence of any obstacles, any path is feasible.

C-obstacle

An obstacle mapped into configuration space. Every obstacle B_i is mapped to the following region in the workspace called the C-obstacle: CB_i = { q in C : A(q) intersected with B_i != empty set }.

C-obstacle region

The union of all the C-obstacles.

Free space

All of the configuration space less the C-obstacle region, called C_free. A configuration in C_free is called a free configuration and a free path is a path where t maps to C_free instead of to C. A semi-free path maps to the closure of C_free.

kinematics

Motions in the abstract, without reference to force or mass. Russell and Norvig define it as the study of the correspondence between the actuator motions in a mechanism and the resulting motion of its parts.

dynamics

Motions of material bodies under the action of forces.

force-compliant motion

Motions in which the robot may touch obstacle surfaces and slide along them. These are more accurate than position-controlled commands.

Rotary motion

Rotation around a fixed hub.

prismatic motion

Linear movement, as with a piston in a cylinder

2007-08-13T05:49:00.000-07:00

Definitions

robot: A versatile mechanical device equipped with actuators and sensors under the control of a computing system. Russell and Norvig define it as "an active, artificial agent whose environment is the physical world."
task planner: A program that converts task-level specifications into manipulator-level specifications. The task planner must have a description of the objects being manipulated, the task environment, the robot, and the initial and desired final states of the environment. The output should be a robot program that converts the initial state into the desired final state.
guarded motion: A motion of the form move along path until sensory-condition.
compliant motion: A motion of the form move along direction d with force = 0 perpendicularly to d. More intuitively, moving along a surface while maintaining a fixed pressure, such as when scraping paint off of a window with a razor.

effector: The bits the robot does stuff with. That is, arms, legs, hands, feet. An end-effector is a functional device attached to the end of a robot arm (e.g., grippers).
actuator: A device that converts software commands into physical motion, typically electric motors or hydraulic or pneumatic cylinders.
degree of freedom: A dimension along which the robot can move itself or some part of itself. Free objects in 3-space have 6 degrees of freedom, three for position and three for orientation.
sensors: Devices that monitor the environment. There are contact sensors (touch and force), and non-contact (e.g., sonar).
sonar: Sensing system that works by measuring the time of flight of a sound pulse to be generated, reach an object, and be reflected back to the sensor. Wide angle but reasonably accurate in depth (the wide angle is the disadvantage).
infrared: Very accurate angular resolution system but terrible in depth measurement.
recognizable set: An envelope of possible configurations a robot may be in at present; recognizable sets are to continuous domains what multiple state sets are to discrete ones. A recognizable set with respect to a sensor reading is the set of all world states in which the robot might be upon receiving that sensor reading.
landmark: An easily recognizable, unique element of the environment that the robot can use to get its bearings.

Motivations

The important incentives for building robots are social, replacing humans in undesirable or dangerous jobs, and economic, reducing the cost of manufacturing while improving its quality.

The real world has the following qualities, that any robot design must take into account:

inaccessible -- sensors are imperfect, and can only perceive local stimuli
nondeterministic -- the robot can never be certain an action will work as expected, since wheels slip, batteries run down, etc.
nonepisodic -- the effects of an action change over time, so robots should handle sequential decision problems and learning
dynamic -- a robot has to know when to think and when to act right away
continuous -- states and actions are drawn from a continuum of physical configurations and motions

2007-08-03T09:48:00.000-07:00

Robotics - Future of Robotics

Artificial intelligence and robotics expert Rod Brooks forecasts major changes in the next 50 years. Much in the way that computers have revolutionized society, robots may take on an increasingly significant role in people's lives. As part of the Gerard Salton Lecture Series, Brooks delivered a talk yesterday entitled "Flesh and Machines: Robots and People" to discuss potential applications of intelligent robots.

Brooks, who directs the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT, asserted that we have more in common with robots, and machines in general, than we think.

"Mankind has had a long history of retreat from 'special-ness,'" Brooks said.

Centuries ago, humans discovered that Earth was not, in fact, the center of the universe. Later, humans and animals were found to have common ancestors. DNA as the fundamental mechanism of life means that humans and yeast are somewhat similar.

"Over time men have become less special and more like technology," Brooks said. "We only have 25,000 genes -- even potatoes have more than that!"

Brooks showed videos of several robots designed in his lab. In one scene, Brooks's colleague Cynthia "plays" with a robot she designed, Kismet.

"We see her moving that eraser, then the robot moving it. They're taking turns." At least, Brooks added, that's what the average observer would think. "But when we thought about it, she was doing all the work. She was giving the robot motion cues. That set us off on reading literature on child development."

Like Cynthia, mothers give their infants motion cues. They engage in activities with their children that the children cannot do by themselves, but can be trained to do with their caregiver's help.

"What the robot sees drives what it does," Brooks said.

Inside these robots exists a three-dimensional space; the robot's emotions are a point in that space. The robot uses its emotional state to generate how it reacts to certain objects, and can display emotion through facial expressions.

In another experiment suggestive of robots' similarity to children in their earliest stages of development, the lab called in various people to speak with the robot.

"When a mother interacts with her child, she generates messages through her voice: praise, attention, prohibition,and soothing are the four basic messages," Brooks said.

In the video, when one woman said, "Good job, Kismet! Look at my smile!" in an encouraging voice, the robot smiled proudly. When another said, "No, no, that's not appropriate" in a disparaging tone, Kismet lowered his head, his large ears drooping.

Although robots like Kismet don't actually understand the meanings of words, they are able to vocally replicate phonemes. As people teach various words to Kismet and Cog, another of CSAIL's robots, the robots can repeat them and identify them with their corresponding objects.

Brooks acknowledges that the development of intelligent robots is still in beginning stages, although significant progress has been made in areas like navigation. However, he said, "I think beyond navigation, robots have new possibilities which will be important."

As the world's demographics shift in the next half century, robots can be useful in fields such as manufacturing, agriculture and elderly assistance. Brooks imagines being able to roboticize large agriculture machines for the maintenance of individual plants. Such robots could do menial and time-consuming tasks like pruning and picking.

"Europe and the U.S. import low-cost labor now ... But that labor may not be there in 50 years," Brooks said.

Second, robot arms could be used for fixed automation, which is particularly useful in manufacturing. Such robots would require the dexterity of a six-year-old, said Brooks. Third, he hoped that robots could be developed to provide in-home care to the elderly, who will soon comprise a much larger demographic in places like North America, Europe, Korea and Japan.

The future, however, holds many challenges to realizing certain robotic applications. "Will we accept robots?" Brooks asked the audience.

It may be hard, he explained, for humans to come to grips with machines that may equal or surpass their own capabilities. Few people want to admit that their emotions can exist within a machine.

"I'm not saying current robots have real emotions, but if they did, it would be hard for people to accept ... and there would certainly be legislation against it!" Brooks said as the audience laughed.

"I liked the lecture very much," said Hugo Fierro grad. "I already took some courses on robots, but never thought about the philosophical aspect of it. I liked his predictions, although they're very futuristic."

"It was a lot of fun. I heard some very interesting and provocative ideas," said Prof. Graeme Bailey, computer science.

And as for the possibility of Brooks' vision becoming reality someday? "I hope so," Bailey said. "If one was to answer no to that, we have a somewhat dismal future for ourselves."

2007-07-17T05:55:00.000-07:00

Robotics - Current Research

The robots of tomorrow will be the direct result of the robotic research projects of today. The goals of most robotic research projects is the advancement of abilities in one or more of the following technological areas:

Artificial intelligence, effectors and mobility,sensor detection and especially robotic vision, and control systems.

These technological advances will lead to improvements and innovations in the application of robotics to industry, medicine, the military, space exploration, underwater exploration, and personal service. The research projects listed below are only a few of many robotic research projects worldwide.

Artificial Intelligence

Cog, a humanoid robot from MIT.

Two of the many research projects of the MIT Artificial Intelligence department include an artificial humanoid called Cog, and his baby brother, Kismet. What the researchers learn while putting the robots together will be shared to speed up development.

Once finished, Cog will have everything except legs, whereas Kismet has only a 3·6-kilogram head that can display a wide variety of emotions. To do this Kismet has been given movable facial features that can express basic emotional states that resemble those of a human infant. Kismet can thus let its "parents" know whether it needs more or less stimulation--an interactive process that the researchers hope will produce an intelligent robot that has some basic "understanding" of the world.

This approach of creating AI by building on basic behaviors through interactive learning contrasts with older methods, in which a computer is loaded with lots of facts about the world in the hope that intelligence will eventually emerge.

Cog is 2 meters tall, complete with arms, hands and all three senses--including touch-sensitive skin. Its makers will eventually try to use the same sort of social interaction as Kismet to help Cog develop intelligence equivalent to that of a two-year-old child.

Kismet is an autonomous robot designed for social interactions with humans and is part of the larger Cog Project. This project focuses not on robot-robot interactions, but rather on the construction of robots that engage in meaningful social exchanges with humans. By doing so, it is possible to have a socially sophisticated human assist the robot in acquiring more sophisticated communication skills and helping it learn the meaning these acts have for others.

Kismet has a repertoire of responses driven by emotive and behavioral systems. The hope is that Kismet will be able to build upon these basic responses after it is switched on or "born", learn all about the world and become intelligent.

Crucial to its drives are the behaviors that Kismet uses to keep its emotional balance. For example, when there are no visual cues to stimulate it, such as a face or toy, it will become increasingly sad and lonely and look for people to play with.

Any advances made with Kismet will be passed on to its big brother Cog, the robot brainchild of Rodney Brooks, head of MIT's AI department.

In mimicking human intelligence, the goal is to make sure robots get a brain and reasoning. An important pioneer in the field of AI is Marvin Minsky.

Without a brain capable of processing input, a robot cannot react to its environment. A brain can be stimulated in hardware or software. Most robots at present have software brains, meaning a computer with a program running. These robots are connected to or equipped with a computer. A drawback is the limited number of processes that can be run on today's computers and the single purpose programs running on these computers. The programs cannot change themselves. In other words, learning is not possible.

A brain made out of hardware, or a number of processors will be closer to reality. The brain will consists of several chips that act both independently and as a group. The general belief is that the real brain works as a neural network of lots of independent processing units. Every chip in itself has a small program. It will process information but also pass it on to other chips. The program changes on a continuous basis. The network of chips is quick and will adapt, so in contrast with the software brain, it will learn.

An example of a hardware brain is Robokoneko the robocat from Genobyte. It has a brain from a machine, the CAM-machine.

Effectors and Mobility

Robot helicopter research began at the University of Southern California in 1991 with the formation of the Autonomous Flying Vehicle Project and continues to the present day. The first robot built was the AFV (Autonomous Flying Vehicle). The AVATAR (Autonomous Vehicle Aerial Tracking And Retrieval), was created in 1994. The current robot, the second generation AVATAR (Autonomous Vehicle Aerial Tracking And Reconnaissance), was developed in 1997. The ``R'' in AVATAR changed to reflect a change in robot capabilities.

Without question, the fish is the best swimmer in the world. That is why the Ship Research Institute of Japan decided to build the Fish Robot. This project hopes to apply what is learned while building and researching with the Fish Robot to the design and construction of ships.

Robots use electro-engines for movement. Engine parts are relatively cheap and last long. Engines are applied to move arm, turn wheels or move other parts, for instance camera's. Engines are less usefull with walking robots. In that particular case engines prove to be a weak part, a jumping robot is a mayor challenge to engine parts. Human being use muscle, which contract and expand, to move around. A muscle receive a signal form the brain and contracts. Causing a joint, like the knee to move.

Material to mimic a muscle is still a dream. Nitinol, an alloy that consist of the metals nickel and titanium will shrink if an electric current travels through the alloy, it will only contract 8% maximum.

The downside, nitinol is very expensive en the contraction is too little to allow it to be used to make walking robots. For the time being walking robots will not use muscles or engines but pneumatic of hydrolic technologies.

To demonstrate advances in research and to stimulate scientist to share progress the Robocup competition is organized a few times a year. Robocup is a competition of Robot soccer teams. Movement, pattern recognition, where's the ball, where's the goal, who is in my team, all this and more is needed to score a goal. A simple games becomes a challenge for a robot team. Besides moving and finding the ball and team members the robots needs to define a strategy and take lots of decisions in a short time frame. Robocup has produced many advancements in both robotic effectors and sensors. Who could have imagined that soccer would contribute to robot research where robots eventually will be smart and capable of cooperation with other to reach a goal?

Sensor Detection

Machine vision involves devices which sense images and processes which interpret the images. Thomas Braunl of The University of Western Australia provides an excellent example of robotic vision research in both hardware and software.

EyeBot is a controller for mobile robots with wheels, walking robots or flying robots. It consists of a powerful 32-Bit microcontroller board with a graphics display and a digital grayscale or color camera. The camera is directly connected to the robot board (no frame grabber). This allows programmers to write powerful robot control programs without a big and heavy computer system and without having to sacrifice vision.

Ideal basis for programming of real time image processing
Integrated digital camera (grayscale or color)
Large graphics display (LCD)
Can be extended with own mechanics and sensors to full mobile robot
Programmed from IBM-PC or Unix workstation,
programs are downloaded via serial line (RS-232) into RAM or Flash-ROM
Programming in C or assembly language

Improv is a tool for basic real time image processing with low resolution, e.g. suitable for mobile robots. It has been developed for PCs with Linux operating system. Improv works with a number of inexpensive low-resolution digital cameras (no framegrabber required), and is available from Joker Robotics. Improv displays the live camera image in the first window, while subsequent image operations can be applied to this image in five more windows. For each sub-window, a sequence of image processing routines may be specified.

Sensor Based Planning incorporates sensor information, reflecting the current state of the environment, into a robot's planning process, as opposed to classical planning , where full knowledge of the world's geometry is assumed to be known prior to the planning event. Sensor based planning is important because: (1) the robot often has no a priori knowledge of the world; (2) the robot may have only a coarse knowledge of the world because of limited memory; (3) the world model is bound to contain inaccuracies which can be overcome with sensor based planning strategies; and (4) the world is subject to unexpected occurrences or rapidly changing situations.

There already exists a large number of classical path planning methods. However, many of these techniques are not amenable to sensor based interpretation. It is not possible to simply add a step to acquire sensory information, and then construct a plan from the acquired model using a classical technique, since the robot needs a path planning strategy in the first place to acquire the world model.

The first principal problem in sensor based motion planning is the find-goal problem. In this problem, the robot seeks to use its on-board sensors to find a collision free path from its current configuration to a goal configuration. In the first variation of the find goal problem, which we term the absolute find-goal problem, the absolute coordinates of the goal configuration are assumed to be known. A second variation on this problem is described below.

The second principal problem in sensor based motion planning is sensor-based exploration, in which a robot is not directed to seek a particular goal in an unknown environment, but is instead directed to explore the apriori unknown environment in such a way as to see all potentially important features. The exploration problem can be motivated by the following application. Imagine that a robot is to explore the interior of a collapsed building, which has crumbled due to an earthquake, in order to search for human survivors. It is clearly impossible to have knowledge of the building's interior geometry prior to the exploration. Thus, the robot must be able to see, with its on-board sensors, all points in the building's interior while following its exploration path. In this way, no potential survivors will be missed by the exploring robot. Algorithms that solve the find-goal problem are not useful for exploration because the location of the ``goal'' (a human survivor in our example) is not known. A second variation on the find-goal problem that is motivated by this scenario and which is an intermediary between the find-goal and exploration problems is the recognizable find-goal problem. In this case, the absolute coordinates of the goal are not known, but it is assumed that the robot can recognize the goal if it becomes with in line of sight. The aim of the recognizable find-goal problem is to explore an unknown environment so as to find a recognizable goal. If the goal is reached before the entire environment is searched, then the search procedure is terminated.

Control Systems

Development of a hierarchical behavior control scheme for rovers and mobile robots is currently underway. It attempts to model and control mobile systems using distinct rule-based controllers and decision-making subsystems that collectively represent a hierarchical decomposition of autonomous vehicle behavior. This research approach employs fuzzy logic, behavior control, and genetic programming as tools for developing autonomous robots. Complex, multi-variable fuzzy rule-based systems are developed in the framework of behavior-based control for autonomous navigation. Genetic programming methods are used to computationally evolve fuzzy coordination rules for low-level motion behaviors. In addition, embedded control applications are being developed for microrover navigation using conventional microprocessors and specialized fuzzy VLSI chips.

The movie Innerspace shows a miniature spaceship travelling through the artery system of a human. It is a nice illustration of the promise of Nano technology. Nano Technology is a technique where miniature robots go to places humans will never be able to travel. Nano technology is a new science where robotics play a mayor part. Questions that needs to be solved because of the very tiny mechinical parts: can a robot repair itself, how do you control a nano robot, how does a nano robot move. Will it be able to work autonomously. Will it be able so shift in shape. Is a nanan robot a mechanical device or is it more like a microprocessor. Once these questions are answered Nano technology will change medical science for ever. Surgery will be performed in lots of cases by one or more Nano robots that will travel inside the human body.

Third generation wireless networks like the Universal Mobile telecommunication System (UMTS) are being developed to support wide band services. A major scenario is to support such services of a user roaming between a cellular terrestrial network and a satellite Personal Communication Network (PCN) while maintaining the quality of service during the hand-over process. And requiring some degree of continuity of quality of service guarantees. This project will focus on developing new protocols which uses artificial intelligent systems to support such hand-over process. Further Seamless roaming and user tracking using intelligent systems will be investigated.

In recent years, the reduction of undesirable vibrations in the dynamic systems such as airplanes, vehicles, tall buildings and off-shore structures has become a crucial issue due to the increased social awareness of comfort as well as the ever increasing heights of new inner city buildings. With the advent of new construction materials and new construction methods, the buildings and structures are becoming taller, and more flexible. With a good design and under normal loading conditions, the response of these structures to vibrations will remain in the safe and comfortable domain. However, there is no guarantee that in-service loads experienced by tall buildings and structures will always be in the allowed range. The undesirable vibration levels could be reached under large environmental loads such as winds and earthquakes, and could adversely affect human comfort and even structural safety. It is becoming critically important to suppress dynamic responses of tall buildings and structures due to the strong winds and earthquakes not only for their safety but also their serviceability. When tall buildings and structures are flexible, design performances may become impossible to achieve by conventional design practice. Hence, additional devices are installed in tall buildings and structures to compensate the dynamic responses caused by environmental loads. As a result, new concepts and methods of structural protection have been proposed.

Due to recent development of sensors and digital control techniques, active control methods of dynamic responses of tall buildings and structures have been developed, and some of them have been implemented to actual buildings. The precondition is however that the implementation is simple enough to be realtime. In engineering applications with rule based systems providing efficient results, the implementation is often easier than its complex conventional counterpart.

The active vibration control in the structural engineering has become known as an area of research in which the vibrations and motions of the tall buildings and structures can be controlled or modified by means of the actions of a control system through some external energy supply. Compared with the passive vibration control the active vibration control can more effectively keep the tall buildings and structures safe and comfortable under the various environmental loads such as strong winds or earthquake hazards. This implies that the active vibration control can be effective and adaptive over a much wider frequency range and also for transient vibration, which is the reason to attract interest of the researchers not only in structural engineering but also in control engineering. Among many methods, that have been proposed, are active mass drivers (AMDs), active tendon systems (ATS), and active variable stiffness systems (AVSs).

Robot manipulators which have more than the minimum number of degrees-of-freedom are termed ``kinematically redundant,'' or simply ``redundant.'' Redundancy in manipulator design has been recognized as a means to improve manipulator performance in complex and unstructured environments. ``Hyper-redundant'' robots have a very large degree of kinematic redundancy, and are analogous in morphology and operation to snakes, elephant trunks, and tentacles. There are a number of very important applications where such robots would be advantageous.

While ``snake-like'' robots have been investigated for nearly 25 years, they have remained a laboratory curiousity. There are a number of reasons for this: (1) previous kinematic modeling techniques have not been particularly efficient or well suited to the needs of hyper-redundant robot task modeling; (2) the mechanical design and implementation of hyper-redundant robots has been perceived as unnecessarily complex; and (3) hyper-redundant robots are not anthropomorphic, and therefore pose interesting programming problems. Our research group has undertaken a broadly based program to overcome the obstacles to practical deployment of hyper-redundant robots.

2007-07-08T21:43:00.000-07:00

Robotics - Interesting points about Robotics

This post is in form of trivia.It deals with some commonly associated questions with robots.

Nowadays, the word robot is often applied to any device that works automatically or by remote control, especially a machine (automaton) that can be programmed to perform tasks normally done by people.

Before the 1960s, robot usually meant a manlike mechanical device (mechanical man or humanoid) capable of performing human tasks or behaving in a human manner. Today robots come in all shapes and sizes, including small robots made of LEGO, and larger wheeled robots that play robot football with a full-size ball.

What many robots have in common is that they perform tasks that are too dull, dirty, delicate or dangerous for people. Usually, we also expect them to be autonomous, that is, to work using their own sensors and intelligence, without the constant need for a human to control them. Looked at this way, a radio controlled aeroplane is not a robot, nor are the radio controlled combat robots that appear on television. However, there is no clear dividing line between fully autonomous robots and human-controlled machines. For example, the robots that perform space missions on planets like Mars may get instructions from humans on Earth, but since it can take about ten minutes for messages to get back and forth, the robot has to be autonomous during that time.

The word robot was introduced in 1920 in a play by Karel Capek called R.U.R. , or Rossum's Universal Robots. Robot comes from the Czech word robota, meaning forced labour or drudgery. In the play, human-like mechanical creatures produced in Rossum's factory are docile slaves. Since they are just machines, the robots are badly treated by humans. One day a misguided scientist gives them emotions, and the robots revolt, kill nearly all humans and take over the world. However, because they are unable to reproduce themselves, the robots are doomed to die. However, the sole surviving human creates a male and a female robot to perpetuate their species.

The roots of robotics can be traced back to Greek mythology and Jewish mysticism. Several myths from Ancient Greece tell of statues being brought to life. According to Aristotle, the legendary Greek inventor, Daedalus (whose son Icarus flew too close to the sun), created animated statues that guarded the entrance to the Labyrinth in Crete. The Jewish Talmud describes the making of a golem, a clay model brought to life by the chanting of magical combinations of letters from the Hebrew alphabet. A similar idea can be found in medieval alchemy, in which the philosopher's stone was believed to have a life-giving force. Mary Shelley drew upon such traditions in her 1818 novel Frankenstein.

The term robotics was coined in the 1940s by science fiction writer Isaac Asimov. In a series of stories and novels, he imagined a world in which mechanical beings were mankind's devoted helpmates. They were constrained to obey what have become known as Asimov's Laws of Robotics:

1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm.

2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Asimov’s book of short stories, I, Robot, investigates the interplay between these laws. In one of the stories, there is a scandal because a candidate for mayor is suspected of being a robot – no-one has ever seen him eat, drink, or sleep.

However, the mayor claims he is not a robot, and the story has many twists and turns before we find out. Is it possible that we would allow robots to run our lives? After all, if they obey the First Law of Robotics, they will never harm us. In this respect they could be better than human politicians! This will not be a practical problem for many years, but who knows what the future holds?

A prototype industrial robot arm named Unimate (designed by George Devol and Joseph Engelberger) was sold to General Motors in 1959. It plucked hot automobile parts out of a die-casting machine and quenched them in water.

The 1960s and 1970s saw a revolution in manufacturing as robots replaced humans for many repetitive jobs. However, these robots were not intelligent by today’s standards. Usually they were programmed by humans training their movements, and they had very little decision-making capabilities. There are still many robots like this in factories today, but the trend is towards more intelligent general-purpose robots that can do more than just paint a panel or screw in a bolt.

Space probes hurtling through the solar system may not seem like robots, but they fully merit that name by performing programmed tasks over long periods without direct human supervision. Operating in the vacuum of space and withstanding exposure to radiation and extremes of temperature, they explore places not yet accessible to humans.

On Christmas Day 2003, the Beagle 2 mission to Mars attempted to land on the Martian surface. Had the landing gone smoothly, the robotic shell would have opened and deployed solar panels to collect electricity, as well as a ‘PAW’ to collect rock and soil samples for a small analytic laboratory. The lander also had a ‘mole’ that would have burrowed into the surface to collect samples for analysis. Beagle 2 is shown on the right with its inventor, Professor Colin Pillinger of the Open University.

It is very difficult to give a robot the ability to perform a wide variety of tasks, move around in cluttered surroundings, recognise objects in the ‘real world’, understand normal speech, and think for itself. These are exciting areas of current research in robotics and artificial intelligence.

For example, the robot shown here has the problem of deciding where to cross the river. How can it make this decision? How would you do it? Perhaps you have come across a similar situation before. Perhaps you could look it up in a guide book. Perhaps you would reason that B is better than C because the water is likely to be shallower? Perhaps you would choose A, because you tried it before. All these ways of making decisions come very naturally to humans, but they are very difficult to program into robots.

Another great problem in robotics is getting them to understand language. This is very important in problem-solving. For example, the four cards below have a letter on one side and a number on the other. If a card has a vowel (a, e, i, o, u) on one side then it has an even number on the other. Which cards do you have to turn over to see if this is true? Think about your answer, then point to a card to turn it over.

Now consider the following cards where the rule is ‘every time I go to Paris I go by plane’. Which cards have to be turned over to test this? Again, think about your answer before turning the card over.

The answer to the first question is that you have to turn over the E to see if it has an even number on the back and you have to turn over the 7 to check that it does not have a vowel on the back. In an experiment, only 12% of people got this second part right (did you?).

The answer to the second question is much easier. Of course you have to turn over the Paris card to check that it has the word plane on the back, but now it’s much more obvious that you have to turn over the train card to make sure it does not have Paris on the back. In the experiment mentioned above, 60% of people got the second part right.

These problems are logically the same, so the experimenters drew the conclusion that the meaning of the symbols is an important part of problem solving. Since robots have very poor language capabilities, their ability to use this kind of reasoning is very limited.

Another of the great problems in robotics is getting them to ‘see’. Although it is easy to put a camera on a robot, it is much more difficult to get the robot to understand what is in an image. Most humans have miraculously good vision. We are able to resolve great ambiguity in scenes. It has proved much more difficult to get robots to understand what is in their universe, and machine vision remains one of the big unsolved problems in robotics research.

There are other problems in robotics that make progress slow. For example, your body is covered with skin, and this contains millions of sensors that allow you to do many fantastically precise things. For example, try typing at a computer with gloves on. The lack of touch feedback will make it very difficult. Also your muscles enable you to have very fine control. Even if you are rather clumsy, you are probably much better at manipulating objects than the average robot. Most people would not let a robot dust their favourite china.

Many futurists believe that robots will eventually and inevitably become more capable than humans, but some experts in artificial intelligence assert that machines will never be able to develop the consciousness and emotions needed for reasoning and creativity.

Nonetheless, there are already commercially available robots that can live in our houses and do basic chores for us. Robots are very good at processing certain kinds of information, and they are ideally suited to answering the telephone and being controlled over the Internet.

The International RoboCup Federation has set itself the challenge of having a team of humanoid robot football players beat the human world champions by 2050. Can you image that? It means that robots will have to become as nimble and skilful as Beckham. It will require the invention of many new materials – for example, a human soccer player could be badly hurt if it clashed with a robot made of metal. It will also require an enormous improvement in machine vision. If you play sports such as football, tennis, or even snooker, next time you play think about the huge amount of information that comes through your eyes.

Robotic pets, lawn mowers and vacuum cleaners are already on the market. Following the success of their Aibo robot dog, Sony have developed a humanoid entertainment robot named QRIO. Honda's Asimo welcomes customers to their showrooms in Japan. Toshiba have built a robot that can play volleyball. Fujitsu's HOAP-2 can perform Japanese Sumo wrestling stances, as well as moves from the Chinese martial art taijiquan.

Rapid advances are being made in robotic control systems, artificial intelligence, neural networks, and in the miniaturisation, sophistication and reliability of electronic circuitry, sensors and actuators. These are all contributing to a steady increase in the capabilities of robots. Robots currently under development may become widely used in the food, clothing, nuclear and offshore industries, healthcare, farming, transportation, mining and defence.

This is a popular science fiction theme, and the answer depends on whether robots will ever attain consciousness and emotions. In stories like 2001: A Space Odyssey and Terminator, humans always find a way to outwit intelligent machines that try to take over control. That's fiction, however, and fact is often stranger than fiction!

The suggestion that robots will take over because they might become more intelligent than humans overlooks one critical fact: the people who have power in human societies are usually not the most intelligent in the obvious, intellectual way. They have different kinds of ‘human intelligence’, including the ability to understand other people, and to influence their behaviour.

The sensible answer to the question as to whether robots will take over is that they probably won’t in the near future. There are many reasons for this. The first is that the robots of today have puny brains compared to humans, and they do not have the ability to organise in the same way as humans. Our societies are very complex and allow us to achieve many very advanced things. It is unlikely that robots could overtake us in the near future. Even so, it is something that we should keep an eye on, since all scientists have a responsibility not to do things that damage society.

However, for the most part, robots play a very positive role in our societies, and we can expect them to be used in many ways that make life better for us all.

If you would like to, why not try out the quiz associated with this minicourse? Go to quiz now.

Alternatively, if this minicourse has taken your interest, why not look at the introductory 10 point credit course, T184 - Robotics and the Meaning of Life

2007-06-30T23:26:00.000-07:00

.TYPES OF ROBOTS

The following article deals with types of robots found in practical applications .It also illustrates how robot should be made inept by giving it the most accurate shape commensurating its operations.

Ask a number of people to describe a robot and most of them will answer they look like a human. Interestingly a robot that looks like a human is probably the most difficult robot to make. Is is usually a waste of time and not the most sensible thing to model a robot after a human being. A robot needs to be above all functional and designed with qualities that suits its primary tasks. It depends on the task at hand whether the robot is big, small, able to move or nailed to the ground. Each and every task means different qualities, form and function, a robot needs to be designed with the task in mind.

Mars Explorer images and other space robot images courtesy of NASA.

Mobile robots are able to move, usually they perform task such as search areas. A prime example is the Mars Explorer, specifically designed to roam the mars surface.

Mobile robots are a great help to such collapsed building for survivors Mobile robots are used for task where people cannot go. Either because it is too dangerous of because people cannot reach the area that needs to be searched.

Mobile robots can be divided in two categories:

Rolling Robots:

Rolling robots have wheels to move around. These are the type of robots that can quickly and easily search move around. However they are only useful in flat areas, rocky terrains give them a hard time. Flat terrains are their territory.

Walking Robots: Robots on legs are usually brought in when the terrain is rocky and difficult to enter with wheels. Robots have a hard time shifting balance and keep them from tumbling. That’s why most robots with have at least 4 of them, usually they have 6 legs or more. Even when they lift one or more legs they still keep their balance. Development of legged robots is often modeled after insects or crawfish..

Robots are not only used to explore areas or imitate a human being. Most robots perform repeating tasks without ever moving an inch. Most robots are ‘working’ in industry settings. Especially dull and repeating tasks are suitable for robots. A robot never grows tired, it will perform its duty day and night without ever complaining. In case the tasks at hand are done, the robots will be reprogrammed to perform other tasks..

Autonomous robots are self supporting or in other words self contained. In a way they rely on their own ‘brains’.

Autonomous robots run a program that give them the opportunity to decide on the action to perform depending on their surroundings. At times these robots even learn new behavior. They start out with a short routine and adapt this routine to be more successful at the task they perform. The most successful routine will be repeated as such their behavior is shaped. Autonomous robots can learn to walk or avoid obstacles they find in their way. Think about a six legged robot, at first the legs move ad random, after a little while the robot adjust its program and performs a pattern which enables it to move in a direction.

An autonomous robot is despite its autonomous not a very clever or intelligent unit. The memory and brain capacity is usually limited, an autonomous robot can be compared to an insect in that respect.

In case a robot needs to perform more complicated yet undetermined tasks an autonomous robot is not the right choice.

Complicated tasks are still best performed by human beings with real brainpower. A person can guide a robot by remote control. A person can perform difficult and usually dangerous tasks without being at the spot where the tasks are performed. To detonate a bomb it is safer to send the robot to the danger area.

Dante 2, a NASA robot designed to explore volcanoes via remote control.

Virtual robots don’t exits in real life. Virtual robots are just programs, building blocks of software inside a computer. A virtual robot can simulate a real robot or just perform a repeating task. A special kind of robot is a robot that searches the world wide web. The Internet has countless robots crawling from site to site. These Web Crawler’s collect information on websites and send this information to the search engines.
Another popular virtual robot is the chatter bot. These robots simulate conversations with users of the Internet. One of the first chatter bots was ELIZA. There are many varieties of chatter bots now, including E.L.V.I.S.

BEAM is short for Biology, Electronics, Aesthetics and Mechanics. BEAM robots are made by hobbyists. BEAM robots can be simple and very suitable for starters.

Robots are often modeled after nature. A lot of BEAM robots look remarkably like insects. Insects are easy to build in mechanical form. Not just the mechanics are in inspiration also the limited behavior can easily be programmed in a limited amount of memory and processing power.

Like all robots they also contain electronics. Without electronic circuits the engines cannot be controlled. Lots of Beam Robots also use solar power as their main source of energy.

A BEAM Robot should look nice and attractive. BEAM robots have no printed circuits with some parts but an appealing and original appearance.

In contrast with expensive big robots BEAM robots are cheap, simple, built out of recycled material and running on solar energy.

Robotics: FAQ

2007-06-22T04:15:00.000-07:00

Robotics long has being the subject of fascination and curiosity,for one and all. From the very advert of smallest of automated machine scientists, technologists and enthusiasts have been steadily working for creating an artificial intelligent which could get as close as possible to humans.

This idea got realized in form of one word "ROBOT".From very beginning it has been marketed as solution to all human woes ,be it frivolous computational problems ,or mundane household chores .Robot has not only lived up to its expectations ,but overwhelmed them to show us new frontiers in fields of space,medicine,industries,research and many more.

But before everyone jumps into this robotics bandwagon which is getting hyped exponentially as seen now-a-days , most of us find it tedious and painstaking to acquire even basics knowledge about it at one place.

Hence ,I have tried to compile some such fundamental queries and their answers
in lucid language ,worth grasping by any lay- man.

Lastly I urge you readers to send comments so that I' m assured of my posts being well- received and my further efforts are in right directions.

1. What is robotics?

Roboticists develop man-made mechanical devices that can move by themselves, whose motion must be modelled, planned, sensed, actuated and controlled, and whose motion behaviour can be influenced by “programming”. Robots are called “intelligent” if they succeed in moving in safe interaction with an unstructured environment, while autonomously achieving their specified tasks.

This definition implies that a device can only be called a “robot” if it contains a movable mechanism, influenced by sensing, planning, actuation and control components. It does not imply that a minimum number of these components must be implemented in software, or be changeable by the “consumer” who uses the device; for example, the motion behaviour can have been hard-wired into the device by the manufacturer.

So, the presented definition, as well as the rest of the material in this part of the post, covers not just “pure” robotics or only “intelligent” robots, but rather the somewhat broader domain of robotics and automation. This includes “dumb” robots such as: metal and woodworking machines, “intelligent” washing machines, dish washers and pool cleaning robots, etc. These examples all have sensing, planning and control, but often not in individually separated components. For example, the sensing and planning behaviour of the pool cleaning robot have been integrated into the mechanical design of the device, by the intelligence of the human developer.

Robotics is, to a very large extent, all about system integration, achieving a task by an actuated mechanical device, via an “intelligent” integration of components, many of which it shares with other domains, such as systems and control, computer science, character animation, machine design, computer vision, artificial intelligence, cognitive science, biomechanics, etc. In addition, the boundaries of robotics cannot be clearly defined, since also its “core” ideas, concepts and algorithms are being applied in an ever increasing number of “external” applications, and, vice versa, core technology from other domains (vision, biology, cognitive science or biomechanics, for example) are becoming crucial components in more and more modern robotic systems.

This part of the post makes an effort to define what exactly is that above-mentioned core material of the robotics domain, and to describe it in a consistent and motivated structure. Nevertheless, this chosen structure is only one of the many possible “views” that one can want to have on the robotics domain.

In the same vein, the above-mentioned “definition” of robotics is not meant to be definitive or final, and it is only used as a rough framework to structure the various chapters in following posts.

This figure depicts the components that are part of all robotic systems. The purpose of this Section is to describe the semantics of the terminology used to classify the chapters in the post: “sensing”, “planning”, “modelling”, “control”, etc.

The real robot is some mechanical device (“mechanism”) that moves around in the environment, and, in doing so, physically interacts with this environment. This interaction involves the exchange of physical energy, in some form or another. Both the robot mechanism and the environment can be the “cause” of the physical interaction through “Actuation”, or experience the “effect” of the interaction, which can be measured through “Sensing”.

Sensing and actuation are the physical ports through which the “Controller” of the robot determines the interaction of its mechanical body with the physical world. As mentioned already before, the controller can, in one extreme, consist of software only, but in the other extreme everything can also be implemented in hardware.

Within the Controller component, several sub-activities are often identified:

Modelling. The input-output relationships of all control components can (but need not) be derived from information that is stored in a model. This model can have many forms: analytical formulas, empirical look-up tables, fuzzy rules, neural networks, etc.

The name “model” often gives rise to heated discussions among different research “schools”, and the post is not interested in taking a stance in this debate: within the post, “model” is to be understood with its minimal semantics: “any information that is used to determine or influence the input-output relationships of components in the Controller.”

The other components discussed below can all have models inside. A “System model” can be used to tie multiple components together, but it is clear that not all robots use a System model. The “Sensing model” and “Actuation model” contain the information with which to transform raw physical data into task-dependent information for the controller, and vice versa.

Planning. This is the activity that predicts the outcome of potential actions, and selects the “best” one. Almost by definition, planning can only be done on the basis of some sort of model.

Regulation. This component processes the outputs of the sensing and planning components, to generate an actuation setpoint. Again, this regulation activity could or could not rely on some sort of (system) model.

The term “control” is often used instead of “regulation”, but it is impossible to clearly identify the domains that use one term or the other. The meaning used in the post will be clear from the context.

The above-mentioned “components” description of a robotic system is to be complemented by a “scale” description, i.e., the following system scales have a large influence on the specific content of the planning, sensing, modelling and control components at one particular scale, and hence also on the corresponding sections of the post.

Mechanical scale. The physical volume of the robot determines to a large extent the limites of what can be done with it. Roughly speaking, a large-scale robot (such as an autonomous container crane or a space shuttle) has different capabilities and control problems than a macro robot (such as an industrial robot arm), a desktop robot (such as those “sumo” robots popular with hobbyists), or milli micro or nano robots.
Spatial scale. There are large differences between robots that act in 1D, 2D, 3D, or 6D (three positions and three orientations).

Time scale. There are large differences between robots that must react within hours, seconds, milliseconds, or microseconds.

Power density scale. A robot must be actuated in order to move, but actuators need space as well as energy, so the ratio between both determines some capabilities of the robot.

System complexity scale. The complexity of a robot system increases with the number of interactions between independent sub-systems, and the control components must adapt to this complexity.

Computational complexity scale. Robot controllers are inevitably running on real-world computing hardware, so they are constrained by the available number of computations, the available communication bandwidth, and the available memory storage.

Obviously, these scale parameters never apply completely independently to the same system. For example, a system that must react at microseconds time scale can not be of macro mechanical scale or involve a high number of communication interactions with subsystems.

Finally, no description of even scientific material is ever fully objective or context-free, in the sense that it is very difficult for contributors to the post to “forget” their background when writing their contribution. In this respect, robotics has, roughly speaking, two faces: (i) the mathematical and engineering face, which is quite “standardized” in the sense that a large consensus exists about the tools and theories to use (“systems theory”), and (ii) the AI face, which is rather poorly standardized, not because of a lack of interest or research efforts, but because of the inherent complexity of “intelligent behaviour.” The terminology and systems-thinking of both backgrounds are significantly different, hence the post will accomodate sections on the same material but written from various perspectives. This is not a “bug”, but a “feature”: having the different views in the context of the same post can only lead to a better mutual understanding and respect.

Research in engineering robotics follows the bottom-up approach: existing and working systems are extended and made more versatile. Research in artificial intelligence robotics is top-down: assuming that a set of low-level primitives is available, how could one apply them in order to increase the “intelligence” of a system. The border between both approaches shifts continuously, as more and more “intelligence” is cast into algorithmic, system-theoretic form. For example, the response of a robot to sensor input was considered “intelligent behaviour” in the late seventies and even early eighties. Hence, it belonged to A.I. Later it was shown that many sensor-based tasks such as surface following or visual tracking could be formulated as control problems with algorithmic solutions. From then on, they did not belong to A.I. any more.

AUTONOMOUS ROBOT

2007-06-15T06:23:00.000-07:00

With obstacle avoidance,Drive wheel synchronization
and line tracking capability

This article is in continuation with my earlier post.So it is expected that u know the basics of designing robot.It is strongly recomended that you go through that post for understanding basics.

The previous post deals with how we can take into account various parameters affecting the motion of robot analytically.Once we design a robot (i.e. fix all dimensions .) , the next question is about powering it .

One way to do this will be have a wired robot with motors attached to wheels.when we supply electric current ,this DC motor will convert that electric energy into mechanical energy and provide the necessary power to robot.The controls will also be handled through this wires.

But more and more application now-a-days require robots which are wire free.To accomplish this we need some power source on board.The only question unanswered is about controls for robot.
This is going to be discussed in this post thoroughly.

This project of ours uses a lot of electronics and electronic components.Hence, any knowledge in this field will be of added importance.Also many pictures and schematic diagrams have been shown along side to supplement the notes and for better understanding.

Project Features:

· ATMega16 Microcontroller
The ATMega16 monitors all robot sensors and
controls motor drive functions.
In system programming is available.
Microcode was developed using the MCS Bascom
AVR compiler.

· Obstacle avoidance
The robot maneuvers around obstacles while
seeking for a black line to track.
Obstacles cause direction reversal when the robot is
in line tracking mode.
The obstacle sensor is a Sharp GP2D15 sensor
connected to INT0 of the ATMega16.

· Line tracking
The robot is designed to search for a black line on
the terrain.
Once the line is found, the robot tracks the line
from end to end and reverses direction when it
encounters an obstacle in the path of the line.
Three infrared sensors mounted under the robot
and connected to port D of the Mega16 do line
sensing.

· Differential Drive
Two DC motors provide mechanical propulsion
through friction drive of the drive wheels.

The motors are electrically driven by two Allegro
3953 full bridge motor drivers connected to port A
of the ATMega16.

Figure 1 - AUTONOMOUS ROBOT DESIGN

· Drive Wheel Synchronization
Infrared wheel encoders and a microcode routine
provide wheel synchronization. This allows for
straight-line travel without guidance.

· Rechargeable Battery Pack
Power is provided by a rechargeable 9.6 Volt NiCad
battery pack.
There is a 5 Volt on-board regulator for logic power.
Project Number A3743
Autonomous Robot
With obstacle avoidance, drive wheel
synchronization and line tracking
capability

AUTONOMOUS ROBOT BRIEF DESCRIPTION

This project is an autonomous robot that is capable of obstacle avoidance, drive wheel
synchronization and line tracking. The robot initializes in seek mode, where it avoids obstacles and searches for a black line to track. Wheel synchronization, during seek mode, allows the robot to travel in a straight line without guidance. Once a black line is found, the robot tracks the line and reverses direction when an obstacle is encountered on the line.

The electronic control assembly consists of an Atmel ATMega16 MCU operating at 8Mhz with an internal clock. Two Allegro 3953 full-bridge motor drivers are used for motor control. Microcode for the project was written using the MCS Basic-AVR compiler. Microcode can be updated using the on-board ISP connector.

The Mechanical assembly uses a differential drive system made from two small DC motors, a friction drive system and two 78mm wheels. There are six on-board sensors that allow the robot to avoid obstacles, synchronize wheel rotation and track lines. Power is provided by a rechargeable 9.6V NiCad battery pack.

Figure 2 – Side View of Robot

AUTONOMOUS ROBOT BLOCK DIAGRAM

AUTONOMOUS ROBOT

AN OVERALL VIEW OF CIRCUITRY

CODE SAMPLE
Microcode for this project was written with the MCS Basic-AVR compiler. This code sample shows the
“Seek_line” routine and drive wheel synchronization method. The robot initializes into the Seek_line
routine, where it roams about looking for a black line to track. The robot must synchronize the two
differential drive wheel rotations, in order to travel in a straight line, when in seek mode. Calling the
“Poll_wheels” routine, where the encoder counters are updated at each wheel encoder pulse and
then comparing the left and right wheel counters does this. The wheel drive is altered if there is a
difference of 2 or more between “Leftc” and “Rightc”.

'* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Seek_line: 'Roam around, avoid obstacles, seek the black line
Leftc = 0 'Initialize wheel counters
Rightc = 0
Motor = Forth 'Go forward
Seek_line1:
Op_mode = "seek"
Gosub Poll_wheels
If Csense = 1 Then 'Center line sensor found black line
Motor = Brake 'Stop
Waitms 300 'pause
Goto Track 'Goto track mode
End If
Motor = Forth 'Give both motors a forward pulse
Gp2 = Rightc + 2 'General Purpose reg = right counter + 2
If Leftc > Gp2 Then 'Is the left wheel 2 pulses ahead of the right wheel?
Motor = Turnl 'Adjust
End If
Gp2 = Leftc + 2 'General Purpose reg = left counter + 2
If Rightc > Gp2 Then 'Is the right wheel 2 pulses ahead of the left wheel?
Motor = Turnr 'Adjust
End If
If Leftc > 130 Then 'Clear the wheel counters after 10 wheel revolutions
Rightc = 0
Leftc = 0
End If
Goto Seek_line1
'* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
'Returns (Leftc) the left wheel encoder count and (Rightc) the Right wheel encoder count.
Poll_wheels: 'Polls the wheel encoder sensors and counts pulses.
If Lwheel = 0 Then 'If left wheel encoder is uncovered
If Left_flag = 1 Then 'Do this once each encoder sensor transition
Incr Leftc 'Increment the wheel counter
Reset Left_flag
End If
End If
If Lwheel = 1 Then Set Left_flag 'Set the flag when the sensor is covered
If Rwheel = 0 Then 'If right wheel encoder is uncovered
If Right_flag = 1 Then 'Do this once each encoder sensor transition
Incr Rightc 'Increment the wheel counter
Reset Right_flag
End If
End If
If Rwheel = 1 Then Set Right_flag 'Set the flag when the sensor is covered
Return

Basics of robotics

2007-06-11T07:38:00.000-07:00

BASICS ON DEVELOPING SIMPLE ROBOT

[1.] Introduction

[2.] The Differential Steering System

[3.] Rounding a Corner

[4.] Some Preliminaries to Developing a Model

[5.] The Derivation

[6.] Dead Reckoning and Odometry

[7.] A Simpler Formula for Dead Reckoning

[8.] Acceleration

Introduction

These notes present equations describing the path of a robot or vehicle equipped with a differentially steered drive system. This simple and reliable wheel-based propulsion system is commonly used in small

ler robots. It is familiar from ordinary life because it is essentially the same system that is used in a wheelchair: two wheels mounted on a single axis are independently powered and controlled,

thus providing both drive and steering functions. The equations in these notes provide a an elementary model for the differentially steered drive system (which is often called a differential steering system). This model can be used to predict how a robot equipped with such a system will respond to changes in its wheel speed and what path it will follow under various conditions. The model can also be used to calculate a robot's position in dead-reckoning or localization by odometry applications (techniques that estimate a robot's position based on distances measured with odometer devices mounted on each wheel).

It is worth emphasizing that the equations given here do represent an elementary model for the motion of a robot or vehicle. They describe the robot's position and orientation as a function of the movement of its wheels, but ignore the underlying physics involved in making that motion happen. Issues such as torques and forces, friction, energy and inertia will be left for a more advanced discussion. In technical terms, this method of describing motion is referred to as a kinematics approach. It ignores the causes of motion (which would be studied in a dynamics approach) and focuses on the effects. These notes also ignore the many interesting details of motors, gearing, electromagnetics, power supplies, and other engineering considerations that make wheel-based actuators possible. As such they provide a first step to understanding the behavior of robots with differentially steered drive systems, not a complete description.

The intended audience for this web page includes high school and undergraduate students who may not be comfortable with the calculus techniques used to develop the differential steering model. Such students should be able to "parse out" the relevant information by scanning the text and paying attention to the results in equations [3] and [5]. To really take advantage of the model requires understanding something of the math and kinematics behind it. So the discussion on this page includes a lot of background information on the problem and a more-than-customary amount of explanation. By providing so much detail, these notes should help the reader apply the differential steering model to his or her own robotic applications.

The differential steering system is sometimes incorrectly called a "differential drive system." The term differential drive is used in automotive engineering to represent a kind of propulsion system that transfers power to its drive wheels through a differential gear or related device (as in the classic rear-wheel drive vehicle). Clearly, it is a different concept than the system described here. Robotics enthusiasts are careful to make the distinction.

The Differential Steering System

As mentioned above, the differentially steered drive system used in many robots is essentially the same arrangement as that used in a wheelchair. So, most of us have an intuitive grasp of the way it behaves. If both drive wheels turn in tandem, the robot moves in a straight line. If one wheel rotates faster than the other, the robot follows a curved path inward toward the slower wheel. If the wheels turn at equal speed, but in opposite directions, the robot pivots. Thus, steering the robot is just a matter of varying the speeds of the drive wheels. But how exactly do we choose the speeds so that the robot will move where we want it to go? In these notes, we will try to refine this intuitive understanding into mathematical expressions that can be used for implementing robot control software.

Rounding a Corner

Before tackling the differential steering model, let's consider the

problem of getting a robot with a to make a 90-degree turn around a corner in a hallway or a road. A simple approach is to have the robot drive to the intersection, stop, and pivot. Clearly, though, such an approach is inefficient and a bit ungainly. A more elegant approach would be for the robot to round the corner, following a gradual circular arc through

the intersection while maintaining a steady speed. To accomplish this end, we increase the speed of the outer wheel while slowing that of the inner.

Is the path that the robot follows truly circular, or is it some other kind of curve? Later on, we will show that if both wheels turn at constant, but different, speeds, the robot does indeed follow a circular path. For now, just assume that path is circular. With that in mind, we see that wheel speeds are selected based on how wide a turn we desire the robot to make.

Figure 1. Path of wheels through a turn.

Referring to Figure 1, observe the relationships:

where sR,sL give the displacement (distance traveled) for the left and right wheels respectively, r is the turn radius for the inner (left) wheel, b is the distance between wheels (from center-to-center along the length of the axle), and theta is the angle of the turn in radians (

).sM is the speed at the center point on the main axle. In this discussion, we will treat the axle's center point as the origin of the simulated robot's frame of reference.

Once we've established the simple geometry for the differential steering system, it is easy to develop algorithms for controlling the robot's path. Note, though, that we did make an important simplifying assumption: the wheels maintain a steady velocity. We neglected the effects of acceleration. If the wheels are allowed to accelerate, the curve which describes the robot's trajectory can become much more complicated. When working with very light robots, where the mass (and inertia) of the platform is small, we can often get away with treating changes in speed as nearly instantaneous. The path that the robot follows will not be truly circular, but it will be close enough for many applications. For larger, heavier robots, of course, mass is important and acceleration must be considered.

Some Preliminaries to Developing a Model

In the discussion above, we considered the problem of trying to choose wheel speeds based on a desired path. Now we will consider the opposite problem: how do we find the trajectory of the robot if we are told what speeds the wheels will be turning? The answer to this question leads to the development of our model and also turns out to be directly applicable to dead reckoning. Essentially, this is a problem in what is called forward kinematics, the technique of predicting a mechanical system's behavior based on the inputs to that system.

Developing a model for the differential steering system is not difficult, but does require the use of calculus and differential equations. If you are unfamiliar with calculus, you may prefer to just view the results given in equations [3] and [5] below.

To begin, consider that what we're really interested in is how x and y coordinates (and orientation) change with respect to time. At any instant, the x and y coordinates of the robot's center point are changing based on its speed and orientation. We treat orientation as an angle q measured in radians, counter-clockwise from the x-axis. Recall that the vector giving the direction of forward motion for the robot will be simply

. The x and y coordinates for the robot's center point will change depending on the speed of its motion along that vector.

These observations suggest that, taking m(t) and theta(t) as time-dependent functions for the robot's speed and orientation, our solution is going to be in the form:

[1]

The Derivation

Let's derive functions for the robot's trajectory. For now, it will be easier to defer thinking about acceleration and just assume that the speeds of the wheels are constant. If both wheels are moving at the same velocity, the robot travels in a straight line and the equation for its trajectory is trivial. So, instead, we will consider the case where the wheels travel at different velocities.

Referring to figure 2, note that as the robot changes position, all points on the robot may be in motion. To develop a forward kinematic equation for the motion of a differential steering system, we start by specifying a frame of reference in which an arbitarily chosen point is treated as stationary. All other points in the system are treated as moving relative to the reference point. The robot is considered a rigid body. By approaching the problem in this manner, we will gain insights which can be applied to the more general problem of modeling the robot's motion in an absolute frame of reference.

Figure 2. Wheels at different velocities.

The point that we select as our reference is the center point of the left wheel. This is the point where an idealized wheel makes contact with the floor. Again, all motion in this frame of reference is treated in relative to the left-wheel point. Because the right wheel is mounted perpendicular to the axle, its motion within the frame of reference follows a circular arc with a radius corresponding to the length of the axle (from hub center to hub center).

Now the central point itself may be in motion, so the actual path of the right wheel will not necessarily correspond to that particular circular arc. But the change in orientation theta is not restricted to the robot's frame of reference. Because we treat the robot as a rigid body, all points in the system undergo the same change in orientation. If we pivot the robot 10 degrees about the left wheel, all points undergo a 10 degree change in orientation. And any change in orientation in the special frame of reference is, in fact, equivalent to that of the more general case.

Based on these observations, we can derive a differential equation describing the change in orientation as with respect to time of time. The definition of an angle given in radians is the length of a circular arc divided by the radius of that circle. The relative velocity of the right wheel gives us that length of arc per unit time. The length from the wheel to the center point gives us the radius. Combining these facts, we have:

[2]

Integrating [2] and taking the initial orientation of the robot as ,

we find a function for calculating the robot's orientation as a function of wheel velocity and time:

[3]

As noted above, this change in orientation also applies to the absolute frame of reference. Shifting our attention back to the absolute frame, we recall from [1] above that the robot's overall motion depends on the the velocity its center point (the midpoint of the axle). That velocity is simply the average of that for the two wheels, or

. We combine this fact with what we know the about orientation as a function of time, and get the following differential equations:

[4]

Note that the equations in [4] are in the same form as that shown in [1]. Integrating and applying the initial position of the robot , we have

[5]

Note that the equations given in [5] confirm the earlier assertion that, when the wheels turn at fixed velocities, the robot follows a circular path. The term

is actually the turn radius for circular trajectory of the robot's center.

When using [5] in a computer application, it is necessary to implement special handling for cases where the wheel speeds are nearly equal and

. In such cases, of course, the robot travels in a nearly straight line. Using L'Hospital's rule from elementary calculus, we can show that the equations do have limits approaching a straight line in the direction of the initial orientation (theta) 0.

Dead Reckoning and Odometry

Once we have [5], it's a small step to be able to perform dead reckoning. In traditional aviation or nautical navigation, the term "dead reckon" means to estimate position without external references. Position is dead reckoned using knowledge about a vehicle's course and speed over a period of time. In robotics, the necessary information is often obtained by measuring wheel revolutions. Devices called encoders are coupled to a robot's drive wheels and act like digital odometers. Although things can go wrong (as when the robot "spins out" on a slippery floor), encoders generally provide a good estimate of displacements for the right and left wheels, respectively. These displacement values can, in turn, be used to determine position based on odometry calculations.

The equations given in [5] can be used for dead reckoning by substituting for the terms and dropping the time value t , then solving for the values x & y, and theta. When using the equations, you need to be careful about cases where the robot travels in a nearly straight-line path and the displacements sL,sR become nearly equal resulting in values near zero in the denominators.

A Simpler Formula for Dead Reckoning

Many popular authors on robotics recommend the formulas shown in [6] below as a way to avoid the complications that we saw in [5].The methods in [6] are especially well suited to small robot applications where on-board computing power is limited (and floating-point operations might not be available).

[6]

The forumulas in [6] are simple and convenient. They work well for algorithms as long as you remember that they are approximations... a fact that is sometimes overlooked. Essentially, [6] computes the robot's total rotation and distance traveled, then treats it as if it had completed the full rotation as a pivot the very beginning of the maneuver and then traveled in a straight line. Of course, this kind of "point-and-shoot" description is an incomplete representation of the reality. But so long as the robot's actual path doesn't involve too much of a turn, everything is fine.

How much of a turn is too much? That depends on the geometry of the particular robot. You can probably use [6] as long as the angle of the turn does not get larger than 10 degrees. If it does, you should apply the approximation in stages. That is, if the robot's transit results in a turn of 20 degrees, treat it as two segments of 10 degrees each.

We proposes a nice little improvement to [6]. Imagine that the robot is following a circular path curving to the left. It is easy to see that the point-and-shoot maneuver described above would create a path to the inside of the curve. Tom suggests that the accuracy of the approximation can be improved by simply dividing the robot's change in orientation by two. So the theta term in the equations becomes

. When adjusted this way, the point-and-shoot path will not be so far inside the curved path and, eventually, the two may even cross. Tom also generously provided an upgrade to the motion appalet to demonstrate the effect of this improvement.

Whether you apply [5] or [6] for your dead reckoning, remember that the realities of mechanical systems limit the accuracy of your calculations. If the robot's wheels hit a slippery spot on the floor and spin wildly, it will degrade the accuracy of the distances measured by the encoders. Even when your wheels maintain traction, your position estimates will be vulnerable to backlash (slop) in the robot's gearing, inaccuracies in the encoders, and small variations in the wheel speed.

Acceleration

Adding acceleration to the model is simply a matter of substituting appropriate time-dependent equations for wheel velocities into [2] and [4]. Unfortunately, the resulting differential equations are often not solvable, but must be evaluated using numerical methods. Let's consider a simple example where the wheels undergo a fixed rate of acceleration. Let

[7]

give the velocities of the right and left wheels where aR,aL give the values for a constant acceleration (or deceleration) and wR,wL give the initial velocities. Substituting into [2] and [4] we have

[8]

We were able to solve the differential equation for orientation directly. The equations for position are intractable and must be evaluated numerically.

In the real world, when a robot applies constant power to a wheel, it does not achieve a constant rate of acceleration. As the parts of the drive system turns faster, it experiences greater internal friction and requires more and more power to realize the same increase in speed. Based on your own system characteristics, you may wish to implement an acceleration model based on a time-dependent polynomial rather than the linear function used in [7]. An even more sophisticated analysis might yield better, more complicated functions for velocity and acceleration. Due to the inaccuracies inherent in robot mechanical systems, however, you will soon reach a point where improving the model is worth neither the human effort or extra CPU processing power needed to do so.

2007-03-12T06:39:00.000-07:00

This bog is dedicated to all those interested in gizmos ,technical, and sci -fi stuff..........
So all interested in this subjects,are more than welcome with all their suggestions...............
To know more just bookmark this page.....................

2007-03-12T00:46:00.000-07:00

hi guys,
this is my first posting. wait and watch for the game!!!!!!!!!!!!!!!!

TECHNOFREAKS

Skeletonization (Roadmap) Methods

Visibility Graphs

Voronoi diagram

Cell Decomposition

Potential Fields

Landmark-Based Navigation

Configuration Space

Generalized Configuration Space

Extensions of the Basic Problem

Multiple Moving Objects

Kinematic Constraints

Uncertainty

Grasp Planning

Computational Complexity

Reducing Complexity

Robot Architectures

Brooks' Subsumption Architecture

Shakey

Situated Automata

Task Planning

Motion Planning

Motion Planning Definitions

Definitions

Motivations

Kismet with its creator, Cybthia Breazeal
of MIT. Breazeal also helped create Cog.

Dante 2, a NASA robot designed to explore volcanoes via remote control.

Robotics: FAQ

AUTONOMOUS ROBOT

Basics of robotics

BASICS ON DEVELOPING SIMPLE ROBOT

Contents

Introduction

The Differential Steering System

Rounding a Corner

Some Preliminaries to Developing a Model

The Derivation

Dead Reckoning and Odometry

A Simpler Formula for Dead Reckoning

Acceleration

TECHNOFREAKS

Kismet with its creator, Cybthia Breazeal of MIT. Breazeal also helped create Cog.

Dante 2, a NASA robot designed to explore volcanoes via remote control.

Robotics: FAQ

AUTONOMOUS ROBOT

Basics of robotics

BASICS ON DEVELOPING SIMPLE ROBOT

Contents

The Differential Steering System

Some Preliminaries to Developing a Model

Dead Reckoning and Odometry

A Simpler Formula for Dead Reckoning

Acceleration

Kismet with its creator, Cybthia Breazeal
of MIT. Breazeal also helped create Cog.