Cooperative Robotics
Transcript of Cooperative Robotics
![Page 1: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/1.jpg)
Cooperative RoboticsLiterature Survey
Presented to the ISRGOctober 24, 2002
Kurt Caviggia
Jonathan Deming
Mazhar Memon
Adam Milner
Nathan Thomas
![Page 2: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/2.jpg)
Presented Research
• Multiple Autonomous Robots (MARS)• Supervising Multiple Autonomous Mobile
Robots • Swarm Intelligence• Cooperative Control of Distributed Multi-
Agent Systems• Multi Robot Coordination for Robust
Exploration
![Page 3: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/3.jpg)
MARS: Multiple Autonomous Robots
University of Pennsylvania
GRASP Laboratory
Philadelphia, Pennsylvania
![Page 4: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/4.jpg)
Current Projects
1. Cooperative Control and Localization of Multiple Robots– Autonomous robots to maintain a formation– Develop a set of graph-based algorithms including:
• Discovery • Cooperative Localization • Cooperative Control
2. Network of Autonomous Robots for Fire-fighting– Cooperative localization and tracking – Conduct search and rescue operations
![Page 5: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/5.jpg)
Robotic PlatformsClodBuster™ I :
1. Camera
2. Wireless Video
3. Micro-controller
ClodBuster™ II :1. Camera
2. On-board PC
ClodBuster™ III :1. Camera
2. Laptop
![Page 6: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/6.jpg)
Cooperative Localization and Control for Multi-Robot
ManipulationJ. Spletzer, A. K. Das, R. Fierro, C. J. Taylor,V. Kumar, and J. P. Ostrowski
GRASP LaboratoryUniversity of Pennsylvania
Philadelphia, PA
Summary:
1. Localization based on visual sensors
2. Control algorithms to maintain a formation
3. A framework for coordinated behavior
![Page 7: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/7.jpg)
Localization by Visualization
Hardware:
•3 Clodbuster™ robots with Omni-directional cameras.
The camera on each robot uses resulting “blobs” frompictures to estimate direction vectors.
Actual Image and “blob” used to estimate direction vectors.
![Page 8: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/8.jpg)
A vector model with Frame 1 as reference.
•Internal Angles are found by using a scalar product.
For Example: ψ2 = cos-1(û21 • û23)
•The sine rule can be applied, giving the position relative to a single robot.
Formation Models
![Page 9: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/9.jpg)
Formations to Coordinate BehaviorCoordinated manipulation :
� Robots surround an object� Leader determines trajectory� Maintain formation
Coordinated Behavior by Formation
![Page 10: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/10.jpg)
Coordination Process
![Page 11: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/11.jpg)
Formation Tests and Results
•Trajectories included both an arc and a circular path. (Left)
•Objects were placed in the center.
•Robots moved an object in “tight” or “loose” formations. (Below)
![Page 12: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/12.jpg)
Applicable Research
• Dynamically centralized control
• Non-global relative localization
• Controlled formation methods
![Page 13: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/13.jpg)
A Hybrid Approach to Supervising Multiple Co-operant Autonomous Mobile
Robots
D P Barnes, R S Aylett, A M Coddington, R A Ghanea-Hercock
UK Robotics Ltd., Manchester, UK
![Page 14: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/14.jpg)
Project Goals
• Control the interaction of robots working together in hazardous conditions.
• Ability to work together to accomplish a specific task.
• Plan, prioritize and coordinate robotic actions to accomplish a task.
![Page 15: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/15.jpg)
System Architecture
• Base station controls planning & organization
• Agents work without low level instructions
• Agents transmit mission status and collected data
![Page 16: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/16.jpg)
Mission Planning Process
![Page 17: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/17.jpg)
Breaking Down Missions
• System Planner Utilizes:– Shape & Dimension of Search Area– Database of each agents capabilities– Algorithms to optimize agent use
![Page 18: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/18.jpg)
Mission Hierarchy
![Page 19: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/19.jpg)
Executing a Behavior
![Page 20: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/20.jpg)
Task Hierarchy
• Agents prioritize their own behaviors• Behaviors are divided into:
– Self Preservation– Environmental Adaptation– Teamwork with other Agents– Action/Task completion
![Page 21: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/21.jpg)
Project Status
• Algorithms tested on 2 mobile robots• The system planner compiles & transmits
missions• Transport, Navigate, Dock, and Track
actions implemented
![Page 22: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/22.jpg)
Useful Information
• System architecture useful for coordinating two or more independent robots
• Defined level of command abstraction• Modular approach to generating instructions• Simple top level communications
– Detailed instructions stored in agents• Method of task prioritization in the mobile
robots
![Page 23: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/23.jpg)
Swarm Intelligence: From Natural to Artificial Systems
Eric Bonaneau, Marco Dorigo, and Guy Theraulaz
Oxford University Press, 1999
![Page 24: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/24.jpg)
Swarm Characteristics
• Large Number of Simple Individuals
• Limited Individual Intelligence
• Complex Coordinated Behavior
• Self Organizing Algorithms
![Page 25: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/25.jpg)
Self Organizing Algorithms
• Stigmergy– Indirect Interactions Through Environment
• Positive Feedback– Recruitment– Time Decay
• Probabilistic
![Page 26: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/26.jpg)
Collection and Sorting
• 3 Rules– Move in Straight Line– Random Turn Away From Obstacle– Drop Object at Group, Then Random Turn
Away• Simple Implementation• Creates Object Clusters• Scales Moderately Well• Efficiency Probabilistic
![Page 27: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/27.jpg)
Collection Diagram
![Page 28: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/28.jpg)
Swarm Mapping
• Rules– Random Path for Individual– Do Not Cross Others Trail– Trail Decays With Time
• Scales Well• Local Optimization Possible• Efficiency Probabilistic• Completeness Probabilistic
![Page 29: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/29.jpg)
Swarm Mapping Diagram
![Page 30: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/30.jpg)
Swarm Mapping Diagram
![Page 31: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/31.jpg)
Advanced Swarm Mapping
• Rules– Divide Area into Subsections– Claim Nearby Unclaimed, Unsearched
Subsection to Search– Search Subsection– Repeat Until Entire Area Searched– Claims Decay With Time
• Scales Well• Efficient• Complete
![Page 32: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/32.jpg)
Advanced Swarm Mapping Diagram
![Page 33: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/33.jpg)
Advanced Swarm Mapping Diagram
![Page 34: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/34.jpg)
Advantages of Swarm Intelligence
• Robust• Flexible• Easily Scalable• Simple Rules• Decentralized• Large Group Memory Available
![Page 35: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/35.jpg)
Disadvantages of Swarm Intelligence
• High Communications Bandwidth• Currently Limited to Simple Tasks• Large Group Memory Needed
![Page 36: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/36.jpg)
Cooperative Control of Distributed Multi-Agent Systems
Marios M. Polycarpou, Yanli Yang, and Kevin M. Passino
Department of Electrical and Computer Engineering and Computer ScienceUniversity of Cincinnati, Cincinnati, OH 45221-0030, USA
Department of Electrical Engineering, The Ohio State University2015 Neil Avenue, Columbus, OH 43210-1272, USA
![Page 37: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/37.jpg)
ProblemCooperative search by a team of distributed agents
Agents:• avoid obstacles or threats• target sensing capabilities• wireless inter-agent communication• computing capabilities to make guidance decisions
![Page 38: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/38.jpg)
Inter- and Outer-loop controllers
Ai - selfV - sensor information from other agentsvi - sensor info from selfP - desired trajectoryui - commands to actuators to trace path
![Page 39: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/39.jpg)
LearningEach agent has a three-dimensional map: z = S(x,y)
z = 1 certain to existz = -1 certain not to existz = 0 total uncertainty
![Page 40: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/40.jpg)
Decision Making• Path Generation - Based on maneuverability
constraints
• Path Selection - Based on cost function
• Cost function - Each agent’s subgoal has an associated cost to penalize or favor a behavior
![Page 41: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/41.jpg)
Path Generation
• A path variation due to maneuverability
• Calculate cost of all subgoals for each generated path
![Page 42: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/42.jpg)
Path SelectionTotal cost: J = w1J1 + w2J2 + w3J3
wx are weights with corresponding costs Jx for subgoal x
Select path with minimum total cost
Possible subgoals:– S1 - Follow path of maximum uncertainty– S2 - Follow path leading to region of maximum uncertainty
– C1 - Follow path of minimum overlap
![Page 43: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/43.jpg)
Artificial potential field method to reduce overlap
Rivaling force is non-zero iff:1) location of Aj is within minimum distance µ and maximum angle φ from
location Ai
2) The difference in heading angle xij(k) between Aj and agent Ai lies within (-x, +x)
![Page 44: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/44.jpg)
Simulation Results2 agents
![Page 45: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/45.jpg)
Simulation Results5 agents
![Page 46: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/46.jpg)
Major contributions
• Creating a straightforward cost based decision making scheme
• “Rivaling Force” to reduce search overlap• Path generation• Path selection
![Page 47: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/47.jpg)
Multi-Robot Collaboration for Robust Exploration
Ioannis Rekleitis
Gregory Dudek
Evangelos Milios
![Page 48: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/48.jpg)
Robot Tracker System
• Localization
• Obstacle/Object Finding
• Exploring
• Adding More Robots
![Page 49: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/49.jpg)
Dead Reckoning• Odometric (Single Robot)
– Optical Encoders• Landmarks (Single Robot)
– Sonar– Laser Range Finders– Assume Complete Information– Optimistic
• Cooperative (Multiple robots)– One Robot Always Stationary– Robot Tracker Sensor– Removes Uncertainty
![Page 50: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/50.jpg)
Cooperative Localization
• Observing Camera
• Accuracy– Few Centimeters– 3 to 5 Degrees
![Page 51: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/51.jpg)
Object and Obstacle Locating• Scans Between
Robots
• Interrupted Visibility Between Robots = Object
• One Sensor type for everything
![Page 52: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/52.jpg)
Exploring Large AreasTrapezoidation Algorithm
1. Break Area into Trapezoids
2. Depth-First Traversal
3. Break Trapezoid into stripes
![Page 53: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/53.jpg)
Motion Strategies
• Motion Strategy A– Straight Lines– Inefficient (D<R)
• Motion Strategy B– Diamond Shape– Optimal (D=R)
![Page 54: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/54.jpg)
Searching With 3 Robots
![Page 55: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/55.jpg)
Searching With 5 Robots
![Page 56: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/56.jpg)
Research Assessment
• Advantages– No Sonar Beacons Required– Sensors Are Not Object Sensitive– No Uncertainties
• Disadvantages– Speed– Can Not Identify What An Object Is– Objects Are Relative to the Robots
![Page 57: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/57.jpg)
Patterned Search Planning and Testing For the Robotic Antarctic
Meteorite Search
Field Robotics Center
Carnegie Mellon University
Pittsburgh, PA
![Page 58: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/58.jpg)
Homogeneous Robot Navigation Planner
• Straight Rows (Back and Forth Pattern)• Spiral Pattern
– Start in Middle following circular pattern– Increase the circles radius every half circle
• Sun-following Pattern– Following nearly a straight row– Slight curve to maintain proper orientation
![Page 59: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/59.jpg)
Simulated Examples of Coverage Patterns
Straight Row Pattern
Spiral Pattern
Sun-Following Pattern
Close-up of Sun-Following Pattern
![Page 60: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/60.jpg)
Navigation Solutions
• Lookahead Distance:– Large
• Gradual and smooth regaining of path• Very time consuming
– Short• Regaining path quicker• May result in oscillations
• Segments– Straight row pattern
• Each row• Each half-circle turn at the ends
– Spiral Pattern• Each half circle
![Page 61: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/61.jpg)
Navigation Solutions
![Page 62: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/62.jpg)
Navigation Information Gained
• Possible Searching Patterns– Straight Row– Spiral – Sun-Following (Not Efficient)
• Look Ahead Distance
![Page 63: Cooperative Robotics](https://reader031.fdocuments.us/reader031/viewer/2022012420/61753557208b2e57745ce0d4/html5/thumbnails/63.jpg)
Thank You for Your Time
Any Questions?