Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

August 31, Reactive Algorithms I

927 views

Published on

Multi-robot Systems

Published in: Technology, Business
  • Be the first to comment

August 31, Reactive Algorithms I

  1. 1. Multi-Robot Systems<br />CSCI 7000-006<br />Monday, August 31, 2009<br />NikolausCorrell<br />
  2. 2. So far<br />Introduction to robotics and multi-robot systems<br />Similar algorithms and properties for robot teams, robot swarms and smart materials<br />
  3. 3. Today<br />Reactive algorithms<br />Environmental templates<br />Collaboration in reactive swarms<br />
  4. 4. Reactive Algorithms<br />Directly couple perception to action<br />Extremely simple hardware (analog electronics will do)<br />Robustness out of simplicity<br />Potential for miniaturization<br />First instance: Grey Walter’s tortoises<br />© The i-Swarm project<br />
  5. 5. Concept: Braitenberg Vehicles<br />Couple perception to action<br />Sensor input coupled to actuator output<br />Inspired by brain architecture <br />left/right hemisphere<br />Neural network<br />Course question: how do the vehicles behave with respect to a light source?<br />Light Sensor<br />Motors<br />
  6. 6. More complex behaviors<br />Braitenberg<br />More sensors (e.g. camera)<br />More connections (e.g. brain)<br />Synthesis by genetic algorithms<br />Modify random connections<br />Unfit individuals fall of the table<br />Hierarchical Decompositon<br />
  7. 7. Subsumption Architecture (Brooks)<br />Decompose behavior into modules<br />Collision avoidance, light following, etc.<br />Arrange modules in layers representing goals<br />Upper layers subsume lower layers<br />Difficult to design with increasing complexity<br />Explore world<br />Wander around<br />Avoid Obstacles<br />Brooks, R. (1986). &quot;A robust layered control system for a mobile robot&quot;. Robotics and Automation, IEEE Journal of 2 (1): 14–23.<br />
  8. 8. Alternative view: Artificial Potential Fields<br />Aka virtual physics, motor schemes<br />Goals are represented by virtual forces (attraction/repulsion)<br /> Forces are calculated from sensor input<br />Addition yields vector field that the robots follow<br />Obvious problem: local minima and cycles<br />© Craig Reynolds<br />
  9. 9. Further Reading<br />ValentionBraitenberg“Experiments in synthetic psychology”, 1986<br />Rodney Brooks“Elephants don’t play chess”, 1990<br />Ronald Arkin“Behavior-based Robotics”, 1998<br />
  10. 10. Example: Jet Turbine Inspection<br />Goal: surround every blade in a turbine with a robotic sensor<br />Robots need to be small, only local communication<br />Alice(ASL, EPFL), sugar cube, 368bytes of RAM<br />
  11. 11. Robotic Platform<br />Alice miniature robot [Caprari2005]<br />PIC microcontroller (368 bytes RAM, 8Kb FLASH)<br />Length of 22mm<br />Maximal speed of 4cm/s, stepper motors<br />4 IR modules serve as very crude proximity sensors (3cm) and local communication devices <br />Energetic autonomy 5h-10h<br />
  12. 12. Baseline: Randomized Coverage without Localization<br />Search<br />Inspect<br />Translate<br />Avoid Obstacle<br />Wall | Robot<br />Obstacle clear<br />Search<br />Inspect<br />Translate<br />along blade<br />pt<br />Blade<br />1-pt<br />Tt expired<br />
  13. 13. Robot Capabilities<br />Sensing: infrared distance sensors<br />Computation: FSM, wall following<br />Actuation: differential wheels<br />Communication: none<br />
  14. 14. Analysis (Intuition)<br />Collaboration: implicit<br />Completeness: probabilistic, asymptotic<br />Probability to leave blade at round or sharp tip affects robot distribution<br />
  15. 15. Experimental Results<br />20, 25, 30 robots<br />
  16. 16. Spatial distribution for pt=0<br />Leaving the blades at a tip generates drift in the environment<br />“Enviromental Template”<br />Probability to inspect some of the blades higher<br />
  17. 17. Exploiting environmental templates: example from Biology<br />Probability to pick up or drop certain objects is a function of local temperature<br />Temperature gradient controls location of objects<br />T<br />3.00 a.m.<br />3.00 p.m.<br />Location of Eggs, Larvae, and Pupae in the nest of the ant Acantholepis Custodiens,<br />© Guy Theraulaz<br />
  18. 18. Randomized Coverage with Collaboration<br />Translate<br />Inspect<br />Inspect<br />Avoid Obstacle<br />Wall | Robot<br />Obstacle clear<br />Search<br />Inspect<br />Mobile<br />Marker<br />pt<br />Blade<br />1-pt | Marker<br />Tt expired<br />
  19. 19. Robot Capabilities<br />Sensing: infrared distance sensors<br />Computation: FSM, wall following<br />Actuation: differential wheels<br />Communication: single bit (blade busy or not)<br />
  20. 20.
  21. 21. Improvement of Collaboration<br />Real<br />Macroscopic Model<br />
  22. 22. Example 2: Stick-Pulling<br />Goal: pull sticks out of the ground<br />Two robots need to collaborate<br />A. Martinoli, K. Easton, and W. Agassounon. Modeling Swarm Robotic Systems: A Case Study in Collaborative Distributed Manipulation. Int. Journal of Robotics Research, 23(4):415-436, 2004.<br />
  23. 23. Robotic Platform<br />16 MHz Motorola CPU<br />Incremental wheel encoders<br />6 frontal infra-red sensors<br />Position feedback in arm (communication!)<br />
  24. 24. Robot Capabilities<br />Sensing: infrared distance sensors, detect stick<br />Computation: FSM, wall following<br />Actuation: differential wheels<br />Communication: explicit, physical via stick<br />Course question: what happens if time-out is too high?<br />
  25. 25. Analysis (Intuition)<br />Time-out during wait key for performance<br />Less robots than sticks<br />Time-out too low: collaboration unlikely<br />Time-out too high: robot depletion<br />More robots than sticks<br />The longer the time-out, the better<br />Optimal value for gripping time when less robots than sticks?<br />
  26. 26. Experimental Results<br />A. Martinoli, K. Easton, and W. Agassounon. Modeling Swarm Robotic Systems: A Case Study in Collaborative Distributed Manipulation. Int. Journal of Robotics Research, 23(4):415-436, 2004.<br />
  27. 27. Example 3: Aggregation<br />Goal: aggregate objects into structures<br />Inspired by nest-building of termites<br />Algorithm<br />Search for seeds<br />Pick-up seed<br />Drop close to other seeds<br />Only seeds at end of cluster are identified as such -&gt; Line formation<br />Martinoli, A., Ijspeert, A.J. and Mondada, F. (1999) Understanding collective aggregation mechanisms: from probabilistic modelling to experiments with real robots. Robotics and Autonomous Systems, 29(1) pp. 51-63.<br />
  28. 28. Aggregation<br />Martinoli, A., Ijspeert, A.J. and Mondada, F. (1999) Understanding collective aggregation mechanisms: from probabilistic modelling to experiments with real robots. Robotics and Autonomous Systems, 29(1) pp. 51-63.<br />
  29. 29. Results<br />Martinoli, A., Ijspeert, A.J. and Mondada, F. (1999) Understanding collective aggregation mechanisms: from probabilistic modelling to experiments with real robots. Robotics and Autonomous Systems, 29(1) pp. 51-63.<br />
  30. 30. Summary<br />Reactive control: tight coupling between perception and actuation<br />Behavior is function of controller and environment<br />Collaboration in reactive swarms<br />Implicit<br />Explicit: via the environment and local communication<br />
  31. 31. Next Sessions<br />Wednesday: More on reactive algorithms<br />threshold-based algorithms<br />message propagation<br />Friday: First lab<br />

×