Mobile Robot Positioning
Multi-Robot Coordination Chapter 11
Objectives Objectives z To understand some of the problems being studied with multiple robots z To understand the challenges involved with coordinating robots z To investigate a simple behaviour-based selforganization strategy for a common application z To investigate a simple communication strategy COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-2
What’s What’s in in Here Here ? z Multi-Robot Coordination: Purpose and Issues Advantages and Disadvantages of Multiple Robots Types of Research and Disciplines Role of Learning z The Foraging Problem What is it ? Explicit Distribution Implicit Distribution Improvement in Distribution z Hierarchical Communication What is it ? Various Schemes - Random - Sequential - Vector - Focused Averaging COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-3
Multi-Robot Coordination: Purpose and Issues COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-4
Multiple Multiple Robots Robots z There are advantages when using multiple robots: larger range of task domains greater efficiency improved system performance fault tolerance lower economic cost ease of development ? distributed sensing and action COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-5
Multiple Multiple Robots Robots z There are also disadvantages / challenges: - performance depends on issues involving interaction between robots - interactions complicate development - difficult to model group behaviors from top down (i.e., centralized control) when environment is unknown and/or dynamic - sensor and/or physical interference - need lots of batteries ! COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-6
Research Research z 5 major themes of robot group research: Group control architecture - decentralization and differentiation Resource conflict resolution - e.g., space sharing Origin of cooperation AAtypical typicalresearch researchpaper paper will willfocus focuson ononly onlyone one theme theme(or (oraspect) aspect)ofof group grouprobotics. robotics. - i.e, genetically-determined social behavior or interaction-based cooperative behavior Learning - e.g., control parameter tuning for desired cooperation Geometric problem solving - e.g., geometric pattern formation COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-7
Research Research z What kinds of problems have been studied: Multi-robot path planning Traffic control Formation generation, keeping and control Target tracking Multi-robot docking Box-pushing Foraging Multi-robot soccer Exploration and localization Transport COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-8
Disciplines Disciplines z There are three disciplines that are most critical to the development of robotic agents: Distributed Artificial Intelligence - distributed Problem Solving or Multi-Agent Systems - considers how tasks can be divided among robots which share knowledge about problem and evolving solutions. Distributed Systems - focus on distributed control addressing deadlock, messagepassing, resource allocation etc Biology - bottom-up approach where robots follow simple reactive rules - Interaction between robots results in complex emergent behavior COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-9
Learning Learning and and Adapting Adapting z Robots perform for certain period of time without human supervision in order to solve problem must be able to deal with dynamic changes in environment and their own performance capabilities z Learning, evolution and adaptation allow robot to improve its likelihood of survival and its task performance in environment: adaptation – how a robot learns by making adjustments learning – helps one robot adapt to environment evolution – helps many robots adapt to environment COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-10
Evolution Evolution vs. vs. Learning Learning z Evolution: process of selective reproduction and substitution based on the existence of a distributed population of vehicles does not perform well when certain environmental changes occur that are different from evolved solutions z Learning: a set of modifications taking place within each individual during its own lifetime often takes place during an initial phase when task performance is considered less important control policy used that gives reasonable performance robot “team” gradually improves over time. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-11
Overview Overview Summary Summary z There are many aspects of multi-robot coordination z Robots that perform well together in one kind of environment will perform poorly in others. z To be useful, multi-robot strategies must: be “designed” and “fine-tuned” for particular applications explicitly / implicitly distribute the work among the robots consider both sensory and environmental interference from other robots be able to operate under unexpected situations be cost-effective COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-12
This This Course Course z Multi-Robot coordination strategies is a huge topic too much to cover in this course z We will consider: self-organization for simple foraging applications hierarchical communication to focus coverage z We will look a simulated results: robots will be reactive and use instinctive behaviors analyze the performance over time combine different types of robots COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-13
The Foraging Problem COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-14
Foraging Foraging z Consider a common problem studied in robotic colonies, foraging: gathering/collecting items - possibly bringing them to some specific location(s) (e.g., to particular room) or general locations(s) (e.g., to outer walls). there are many variations of this problem z We will consider a specific instance: robots can detect when it finds an item and can push it to some location (or pick it up and drop it off). robots will be encoded with a fixed, instinctive behavior and thus will not learn “how” to forage. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-15
Foraging Foraging z Consider allowing robots to move randomly in an environment with no cooperation. z Robots must find forage items (e.g., when passing over them) and bring them to the boundaries. Robots may collide, which may interrupt the forage procedure of a robot. Eventually, over time, each forage item will be found by some robot: COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-16
Foraging Foraging 100 90 5 robots 80 10 robots 70 % Completion z As more robots are used, the speed of forage completion increases. Foraging Performance Over Time - Random Movement with Evenly Spread Forage Items 25 robots 50 robots 60 100 robots 50 200 robots 1 robot 40 30 z The performance decreases when the forage items are not evenly distributed. 20 10 0 Increasing Time Foraging Performance Over Time - Random Movement with Clustered Forage Items 100 90 COMP 4900A - Fall 2006 10 robots 70 % Completion this is because robots are not directed towards forage items, only finding them by chance. 5 robots 80 25 robots 50 robots 60 100 robots 50 200 robots 1 robot 40 30 20 10 0 Chapter 11 – Multi-Robot Coordination Increasing Time 11-17
Foraging Foraging z Intuitively, performance can be improved by: reducing collisions (or interference) between robots preventing robots from traveling over the same areas directing robots towards clusters of forage items z The obvious way of reducing collisions and preventing duplicate travel is to distribute robots by explicitly assigning each one a particular area in the environment in which to forage. environment broken down into “equal-sized” areas which are assigned to individual robots COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-18
Foraging Foraging –– Explicit Explicit Distribution Distribution z This strategy has advantages: ensure even distribution of robots - good when items to be foraged are evenly distributed randomly minimizes sensor interference and physical collisions between robots z and disadvantages: - requires robots to “know” and maintain specific positions - requires knowledge of environment - expensive sensors ? (e.g., GPS) - expensive computation (e.g., position estimation) - can be inefficient if forage items are clustered COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-19
Foraging Foraging –– Explicit Explicit Distribution Distribution z A simple way of determining the foraging areas for each robot is to base the regions on the dual graph: Recursively divide dual graph in “half” until number of regions matches the number of robots: COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination Each Eachrobot robot remains remainsinin itsitsown own designated designated area. area. 11-20
Foraging Foraging –– Explicit Explicit Distribution Distribution z There are multiple ways to split the dual graph by finding an edge that evenly splits: links – # of dual graph links - simple and fast, assuming a nice triangulation area – area covered by dual graph triangles - best if robots need to perform coverage algorithms or searching with uniform distribution of foraging items. perimeter – perimeters of dual graph triangles - good if robots are to patrol outer boundaries of their environment COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-21
Foraging Foraging –– Explicit Explicit Distribution Distribution z Performance (i.e., speed of forage completion) is highly dependant on shape of environment and location of forage items. With Withforage forageitems itemsevenly evenlydistributed, distributed,robots robotswork workeffectively effectivelyinin near optimal configuration, provided that robots do not have to near optimal configuration, provided that robots do not have to leave leavetheir theirenvironment environmenttotocomplete completethe thetask. task. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination With Withclustered clusteredforage forageitems, items, most robots become useless most robots become useless ififforced forcedtotoremain remainininaa particular area. particular area. 11-22
Foraging Foraging –– Implicit Implicit Distribution Distribution z Clearly, fixing the locations of each robot may not be the best choice if: the distribution of forage items is not known to be random and evenly distributed the robots must travel outside their areas to complete the forage task (i.e., to deliver their payload). z A compromise is to hard-code specific behavioral rules into the robots that minimize their collisions and attempt to keep them distributed. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-23
Foraging Foraging –– Implicit Implicit Distribution Distribution z Consider robots with omni-directional beacons which are detectable from other nearby robots: robots avoid moving towards nearby beacons intuitively, robots should remain separated/distributed When other robot detected When other robot detected within sensor range, robot within sensor range, robot moves in opposite direction. moves in opposite direction. With multiple beacons, With multiple beacons, either move away in either move away in combined vector combined vector direction or away from direction or away from strongest signal. strongest signal. COMP 4900A - Fall 2006 Although robots may still reAlthough robots may still reencounter other robots during encounter other robots during their movements, in general their movements, in general they remain distributed. they remain distributed. Chapter 11 – Multi-Robot Coordination 11-24
Foraging Foraging –– Comparison Comparison z A comparison of these schemes shows that: for evenly spread forage items there is no significant advantage of either scheme in terms of forage completion time and the simple random movement seems to do well. for clustered forage items the fixed area scheme performs poorly with few robots and the repel scheme performs better Scheme Comparison - 25/12/4 Robots Evenly Spread Forage Items Repel scheme favorable Repel scheme favorable since performs well AND since performs well AND minimizes robot contact. minimizes robot contact. 100 90 80 70 60 80 Repel (12 robots) Fixed (12 robots) Random (12 robots) 50 40 30 20 Repel (4 robots) Fixed (4 robots) 10 0 % Completion % Completion 100 90 Scheme Comparison - 25/12/4 Robots Clustered Forage Items 70 Repel (12 robots) 60 Fixed (12 robots) 50 Random (12 robots) Fixed (4 robots) Random (4 robots) 30 Repel (25 robots) 20 Repel (25 robots) Fixed (25 robots) 10 Fixed (25 robots) Random (25 robots) Random (4 robots) 0 Increasing Time COMP 4900A - Fall 2006 Repel (4 robots) 40 Random (25 robots) Increasing Time Chapter 11 – Multi-Robot Coordination 11-25
Foraging Foraging –– Improvement Improvement z A more significant improvement can be made if something is known about the forage items (e.g., they are clustered). can “signal” other robots when item is encountered Robot turns on beacon Robot turns on beacon when item is found. when item is found. leave signal on until: - fixed amount of time elapses - other robots come nearby Robots within beacon’s Robots within beacon’s range will travel toward range will travel toward nearest beacon. nearest beacon. Robots outside of Robots outside of beacon’s range will beacon’s range will continue moving continue moving randomly. randomly. can either wait stationary or continue moving COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-26
Foraging Foraging –– Improvement Improvement z Consider five “beacon attraction” schemes: Always On - beacon is always on, robot keeps moving Timed Out Stationary - beacon on for fixed time, robot waits stationary until beacon timeout Timed Out Moving - beacon on for fixed time, robot keeps moving Until Near Robots Robotsmay mayget getinto into aadeadlock situation. deadlock situation. - beacon on until robot nearby, robot waits stationary until another robot comes nearby Until Near or Timed Out - beacon on for fixed time, robot waits stationary until beacon timeout or until another robot comes nearby COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-27
Foraging Foraging –– Improvement Improvement z Here is the basic idea behind the attraction code: REPEAT { int desiredDirection direction of closest/strongest beacon signal; IF (desiredDirection ! null) { boolean collisionDetected read front collision sensors; IF (collsionDetected) { Turn away from obstacle; Depends Dependson onsensor. sensor. The The } desired direction may be that desired direction may be that Turn towards desiredDirection ofofthe thestrongest strongestsignal signal(if(ifmany many } beacons sensors are mounted beacons sensors are mounted ELSE { ininaacircular circularfashion), fashion),orormay may wander (i.e., move forward or turn randomly) be a direction representing a be a direction representing a } combination combinationofofmultiple multiple } signals. signals. Usually, Usually,the thedirection direction will be one of 8 to 16 fixed will be one of 8 to 16 fixed directions directionsaround aroundthe therobot. robot. IF (a forage item is found) { IF (a forage item is found) { Turn on my beacon; Turn on beacon; Wait for XXX seconds; counter 5000; //msec Turn off my beacon; } } IF (--counter 0) { Turn off beacon; Add this code for the Add this code for the TimedOutStationary scheme TimedOutStationary scheme COMP 4900A - Fall 2006 } Add Addthis thiscode codeinstead insteadfor forthe theTimedOutMoving TimedOutMovingscheme scheme Chapter 11 – Multi-Robot Coordination 11-28
Foraging Foraging –– Improvement Improvement z What about performance ? Scheme Comparison - 12 Robots Clustered Forage Items Movement Movement while whilebeacon beacon isison, is on, isbest best strategy. strategy.Up Up toto3x faster 3x faster than thanrandom random movement movement here. here. % Completion Attract & Move Until Timeout Attract & Still Until Timeout No Attract (Random) 100 90 80 70 60 50 40 30 20 10 0 Attract Until Near or Timeout Attract Always Attraction Attractionwith withtimed timedbeacon beacon always improves performance. always improves performance. Attracting Attractingallallthe thetime, time,can canbe be worse worsethan thanmoving movingrandomly. randomly. Increasing Time COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-29
Foraging Foraging –– Improvement Improvement z Even when varying the number of robots, the attraction scheme performs well: Scheme Comparison - 100 Robots Clustered Forage Items Scheme Comparison - 4 Robots Clustered Forage Items 100 100 % Completion 80 70 60 90 Attract Until Near or Timeout 70 Attract & Still Until Timeout 50 40 Attract Always 30 20 Attract & Move Until Timeout 80 % Completion 90 Attract & Move Until Timeout Attract Until Near or Timeout 60 Attract & Still Until Timeout 50 40 Attract Always 30 20 No Attract (Random) 10 0 No Attract (Random) 10 0 Increasing Time Increasing Time The Thetime timescales scalesbetween betweenthe thegraphs graphsisis different, different,ininorder ordertotoaccentuate accentuatethe the differences in the schemes. differences in the schemes. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-30
Foraging Foraging –– Improvement Improvement z Of course, in non-clustered environments, the attraction scheme performance degrades and actually reduces efficiency over random scheme: Scheme Comparison - 12 Robots Evenly Spread Out Forage Items Recall Recallthat thatrepel repel scheme works scheme worksbest best ininunclustered unclustered environments. environments. % Completion Attract & Move Until Timeout Attract & Still Until Timeout No Attract (Random) 100 90 80 70 60 50 40 30 20 10 0 Attract Until Near or Timeout Attract Always Repel Always Attraction Attractionschemes schemesperform performworse worse than simple random movement. than simple random movement. Increasing Time COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-31
Foraging Foraging –– Improvement Improvement z What about environments with both clustered items AND spread out items ? Scheme Comparison - 12 Robots Clustered AND Evenly Spread Out Forage Items % Completion Attract & Move Until Timeout Attract & Still Until Timeout No Attract (Random) Attract Until Near or Timeout Attract Always Repel Always 100 90 80 70 60 50 40 30 20 10 0 Performance Performanceisisnear neartoto random random but butprovides providesonly only aasmall small improvement. improvement. Increasing Time COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-32
Foraging Foraging –– Improvement Improvement z Can mix various kinds of robots: e.g., some attract, some repel Scheme Comparison - 12 Robots Clustered AND Evenly Spread Out Forage Items % Completion No Attract (Random) Random AND Attract Near Attract Until Near or Timeout Random AND Attract Move Attract & Move Until Timeout Combining Combining66random random with 6 attract robots with 6 attract robots performs performsbest bestdespite despite type of environment type of environment!!!! 100 90 80 70 60 50 40 30 20 10 0 Increasing Time Scheme Comparison - 12 Robots Clustered Forage Items Scheme Comparison - 12 Robots Evenly Spread Out Forage Items Random AND Attract Move Attract & Move Until Timeout 100 90 80 70 60 50 40 30 20 10 0 No Attract (Random) Random AND Attract Near Attract Until Near or Timeout % Completion % Completion No Attract (Random) Random AND Attract Near Attract Until Near or Timeout Increasing Time COMP 4900A - Fall 2006 Random AND Attract Move Attract & Move Until Timeout 100 90 80 70 60 50 40 30 20 10 0 Certainly, Certainly,less lessrobots robotsare areattracted attractedtoto cluster, so in clustered environment, cluster, so in clustered environment, there thereisisaaperformance performancetradeoff tradeoffwhen when combining robot types. combining robot types. Increasing Time Chapter 11 – Multi-Robot Coordination 11-33
Other Other Similar Similar Problems Problems z Similar attraction/repel strategies can be implemented for other problem scenarios such as coordinated mapping, searching, patrolling, floor cleaning etc. same principles apply, but results may differ. z As seen, using heterogeneous groups (i.e., mixing different kinds of robots) may prove to be the most robust and efficient solution overall. z Experimentation helps to tweak solutions: wanna do an honours project or a Master’s thesis ? COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-34
Hierarchical Communication COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-35
Communication Communication z Another important issue with respect to multi-robot algorithms has to deal with communications: do the robots need to communicate (e.g., send data) ? is there any advantage to doing so ? how often should they communicate ? should there be unlimited communication between robots or should there be restrictions (i.e., groups) ? z We will look here at one aspect of using hierarchical communication. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-36
Hierarchical Hierarchical Communication Communication z Consider robots organized into a hierarchy: Each robot belongs to a group and all group members can communicate to a group “leader” via wireless communication. The leaders are also grouped together with a higher level leader to which they communicate. COMP 4900A - Fall 2006 High level leader communicates High level leader communicates with 3 middle level leaders. with 3 middle level leaders. 55Low Lowlevel levelworker workerrobots robots communicate communicatewith withtheir their leader leaderas aslong longas asthey theyare are within withincommunication communicationrange. range. Chapter 11 – Multi-Robot Coordination 11-37
Hierarchical Hierarchical Communication Communication z Within a hierarchy, worker robots must always remain within communication range: allows data to be transmitted to leader (e.g., map data) allows leader to send commands at any time (e.g., new Warning directions and updated task assignments) Warning Communication Communicationrange rangelimit limit zone zone allows quick docking for battery recharging, working in shifts etc an warning buffer zone should be used to inform worker to turn back. Almost Almostout outofofrange, range, needs needstototurn turnback. back. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-38
Hierarchical Hierarchical Communication Communication z A main issue with bottom-up behavior-based programming is that only local information (i.e., information from a robot’s own sensors) is usually available. z With such a hierarchical scheme, lower level robots can be given global knowledge of the environment and/or of task completion. should provide benefit over no-communication schemes for more complex problems can allow “steering” of robots to accomplish task more efficiently. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-39
Hierarchical Hierarchical Schemes Schemes z Consider robots moving randomly to cover a simple environment: good enough to investigate the general problem of robot coverage under various communication schemes. more efficient schemes can be used to cover environment and techniques can be “tweaked” to each application. Random coverage of Random coverage of 4 robots over time. 4 robots over time. random coverage actually performs well over time. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-40
Hierarchical Hierarchical Schemes Schemes z Now consider a leader with 4 worker robots: worker robots move randomly within leader’s communication range: Robots Robotsallallmove move randomly within randomly within communication communication range. range. we can restrict worker movements to fixed or variable-sized wedges/quadrants: Robots Robotsmay maycross crossover overinto intoother other quadrants, quadrants,but buttreat treatititas asout outofofrange. range. COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-41
Hierarchical Hierarchical Schemes Schemes z Leader must also move in order to cover whole environment properly. z Consider various leader movement schemes: - Random: move in random direction - Sequential: move along a fixed path in sequence - Vector: move in direction towards quadrant that had most “out of safe zone” occurrences - Toward Average: vector scheme with added “pull” towards leader’s average location - Away From Average: vector scheme with added “push” away from leader’s average location COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 11-42
Hierarchical Hierarchical Scheme Scheme -- Sequential Sequential z The basic sequential scheme works as follows: Leader moves slower than workers (e.g., 1/10th of speed) Leader heads towards next location in some sequence (e.g., along a predetermined path) Leader may remain at each location for a while or leave immediately. Timeout may be used if location is not reached within certain time limit COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination 4 5 3 2 1 Leader Leadermoves movesalong alongpath, path, while whileworkers workersmove moverandomly randomly within withinthe the“safe” “safe”range. range. Good Goodififneed needtotounload unloadworkers, workers,then then reload and transport to new site. reload and transport to new site. Necessary Necessaryininorder ordertoto avoid avoidgetting gettingstuck stuck behind obstacles. behind obstacles. 11-43
Hierarchical Hierarchical Scheme Scheme -- Vector Vector z The basic vector scheme works as follows: Leader moves slower than workers (e.g., 1/10th of speed) Each time worker leaves “safe” range, a counter is incremented 2 1 1 2 5 4 3 3 1 3 1 2 Leader computes 4 vectors facing 4 quadrants with magnitudes equal to these counters 5 Leader Leadermoves moves inincombined combined vector vectordirection. direction. 3 Leader moves - in combined direction of these vectors, or - in direction of strongest magnitude vector COMP 4900A - Fall 2006 2 Chapter 11 – Multi-Robot Coordination 2 3 11-44
Hierarchical Hierarchical Scheme Scheme –– Average Average Vector Vector z The average vector scheme works as follows: Same 4 vectors as Vector scheme are used Leader also keeps track of its overall average position Leader computes 1 new vector facing either towards or away from the global average according to its current location Average Average Leader includes this new vector in its computations Magnitude of global average vector set to scalar multiple of maximum of other vectors (e.g., 2x, 1x, ½x, etc ) COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination vector, vector, set settoto1x 1x maximum maximum 5 3 2 3 5 11-45
Hierarchical Hierarchical Results Results z Results from the Random movement scheme: Leader Leadermoves movesrandomly, randomly, while workers stay while workers staynearby. nearby. This Thisstrategy strategymay maynot notreach reach allallparts of the environment. parts of the environment. Combined Paths of Workers COMP 4900A - Fall 2006 Too much Too much clustering clustering on onedges. edges. Chapter 11 – Multi-Robot Coordination Leader’s Path 11-46
Hierarchical Hierarchical Results Results z Results from the 4-Point Sequential scheme: Leader Leadermoves movestotopredefined predefinedlocations locations(in (inthis this case, case,44“corners”), “corners”),while whileworkers workersstay staynearby. nearby. Nice coverage in general, but corners Nice coverage in general, but corners are missed. How can we fix this ? are missed. How can we fix this ? Combined Paths of Workers COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination Leader’s Path 11-47
Hierarchical Hierarchical Results Results z Results from the Vector scheme: Performs Performsok, ok,but butnot notbetter better than without communication. than without communication. Leader Leadermoves movestoward towarddirection directionofofworker worker that was out of safe range the most that was out of safe range the mosttimes. times. Combined Paths of Workers COMP 4900A - Fall 2006 Chapter 11 – Multi-Robot Coordination Leader’s Path 11-48
Hierarchical Hierarchical Results Results z Results from the Toward Average Vector scheme: good for applications such as focused searching in which the likelihood of success is localized about Can Canform formsearch search“rings” “rings”by by some known location. Can Cankeep keepless lessfocus focustoto varying allow outward expansion. varyingmagnitude magnitudeover overtime. time. Can Canreally reallyfocus focusattention attentionofof workers workersaround aroundaaspecific specificarea. area. 2x Attraction Magnitude COMP 4900A - Fall 2006 allow outward expansion. 1x Attraction Magnitude Chapter 11 – Multi-Robot Coordination ½ x Attraction Magnitude 11-49
Hierarchical Hierarchical Results Results z Results from Away From Average Vector scheme: good for applications such as mapping to “force” exploration away from previously mapped areas. Can Canreally reallyfocus focusattention attentionofof workers away from workers away fromaaspecific specificarea. area. 2x Repel Magnitude COMP 4900A - Fall 2006 Can Cankeep keepless lessfocus focustoto allow inward expansion. allow inward expansion. 1x Repel Magnitude Chapter 11 – Multi-Robot Coordination Can Canuse useaa“hint” “hint”ofoffocus
50 robots 100 robots 200 robots 1 robot Foraging Performance Over Time - Random Movement with Clustered Forage Items 0 10 20 30 40 50 60 70 80 90 100 Increasing Time % Completion 5 robots 10 robots 25 robots 50 robots 100 robots 200 robots 1 robot. COMP 4900A - Fall 2006 Chapter 11 - Multi-Robot Coordination 11-18
As shown in Figure 1, positioning systems can be classified into two categories: 1) Global Positioning 2) Local Positioning Global Positioning Systems (GPS) allow each mobile to find its own position on the globe. Local Positioning System (LPS) is a relative positioning system and can be classified into Self and Remote Positioning.
MELSEC iQ-F FX5 User's Manual (Positioning Control) 7 FX Series Programmable Controllers Introduction to FX Positioning Control Systems 1 The Basics of Positioning Control 1.1 What is positioning control? 1 The Basics of Positioning Control 2 Positioning by AC Servo System 3 Components of Positioning Control 4
robot - kuka kr iontec p. 8/9 range of electrospindles for industrial robots gamma di elettromandrini per robot industriali p. 6/7 robot - kuka kr quantec p. 12/13 robot - kuka kr quantec robot - kuka kr 360 fortec robot - kuka kr 500 fortec robot - kuka kr 600 fortec p. 16/17 rotary tables tavole rotanti p. 20/21
steered robot without explicit model of the robot dynamics. In this paper first, a detailed nonlinear dynamics model of the omni-directional robot is presented, in which both the motor dynamics and robot nonlinear motion dynamics are considered. Instead of combining the robot kinematics and dynamics together as in [6-8,14], the robot model is
the robot's true position. Index Terms—Position, Mobile Robot, Extended Kalman Filter, Simulation. I. II. INTRODUCTION To track a mobile robot movement in a 2D. The position of the robot (x and y) is assumed to be sensed by some sort of GPS sensor with a suitable exactitude, while the angle orientation of the robot will be acquired by applying
1. The robot waits five seconds before starting the program. 2. The robot barks like a dog. 3. The robot moves forward for 3 seconds at 80% power. 4. The robot stops and waits for you to press the touch sensor. 5. The robot moves backwards four tire rotations. 6. The robot moves forward and uses the touch sensor to hit an obstacle (youth can .
In order to explore the effect of robot types and task types on people s perception of a robot, we executed a 3 (robot types: autonomous robot vs. telepresence robot vs. human) x 2 (task types: objective task vs. subjective task) mixed-participants experiment. Human condition in the robot types was the control variable in this experiment.
Positioning Yourself and Your Career You can benefit by using positioning strategy to advance your own career. Key principle: Don't try to do everything yourself. Find a horse to ride Chapter 24. Positioning Your Business To get started on a positioning program, there are six questions you can ask yourself Chapter 25. Playing the Positioning Game
