1. Chapter 40 Springer Handbook of Robotics, ©2008 Presented by:Shawn Kristek

2. 1. Task Complexity 2. Distributed tasks 3. Difficult to build ultimate robot 4. Parallelism 5. Robustness through redundancy

3.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

4.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

5. o Centralized o Hierarchical o Decentralized o Hybrid

6. Centralized  Single point of control Works best when controller oversees other robots X Vulnerable to single robot failure X Real-time difficulties due to communication requirements http://rsl.engr.scu.edu/NewWeb/News/imagesSpr04/2004op2.jpg

7. Hierarchical  Similar to military command More resistant to single robot failures X Vulnerable to upper-level single robot failure

8. Decentralized  Most common  Each robot’s actions based on localized data Robust to single robot failure X Global coherency difficult -Incorporated high-level goals difficult to revise Matarić http://www.dailyspeculations.com/wordpress/?p=1847

9. Hybrid  Combinations of other architectures Advantages of levels of control and localized control -Robust to failures -Global coherency DIRA

10. Implementations  The NERD Herd  ALLIANCE  DIRA

11. Implementations: The NERD Herd Matarić  Decentralized  Stigmergic  Swarm robots  Homogeneous: 20 identical robots  Behavior-based  Applications -Foraging & Coverage -Flocking & Formations The NERD Herd

12. Implementations: ALLIANCE Parker  Decentralized  Minimal explicit communication  Heterogeneity possible  Behavior-based - Uses motivations  Applications -Box pushing & Cooperative Manipulation -Multitarget observation ALLIANCE

13. Implementations: DIRA DI stributed R obot A rchitecture Simmons  Hybrid  Explicit communication  Heterogeneity possible  Applications - Cooperative Manipulation DIRA

14.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

15. o Stigmergy o Passive action recognition o Explicit

16. Stigmergy  Sense through the world Simple No communication channels & protocols X Limited by robot’s perception Melhuish and Holland

17. Passive action recognition  Communication through observation No limited bandwidth No fallible mechanism X Limited by robot’s perception X Difficult to analyze actions

18. Explicit  Direct communication -Synchronize actions -Exchange information -Negotiate Directness Ease of acquiring knowledge of teammates X Noisy, limited-bandwidth channel

19.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

20. o Swarm Robots – Homogeneous o Heterogeneous

21. Swarm Robotics Collective robotics  Typically homogeneous  Biologically inspired -Ants -Bees  Stigmergic  Redundant

22. Heterogeneous More realistic: - Heterogeneity may emerge in homogeneous systems Provides various capabilities Can reduce costs X Unavoidable Parker Grabowski

23.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

24. o Taxonomy o Approaches

25. What is it?  Efficiently assign tasks  Team goal defined as set of tasks  Each task can be subdivided Goal

26. Taxonomy Gerkey and Matarić  Tasks -SR : Single-robot task -MR: Multirobot task  Robots -ST: Single-task robot -MT: Multitask robot  Allocation Optimization - IA: Instantaneous Assignment -TA: Time-extended Assignment  SR-ST-IA

27. Approaches  Behavior-Based  Market-Based

28. Behavior-Based  Decentralized architecture  Avoids explicit communication  Task Allocation -Current state -Teammate capabilities

29. Market-Based  Negotiation/bidding based  Greedily assigned to robot with highest utility  Most focus on SR-ST-IA/TA  Centralized or Hybrid architecture  Explicit communication  M+ architecture of Botelho and Alami -First for multirobot -Individual plans merged for team benefit

30.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

31.  Many difficulties -Exponential state spaces -Limited training time -Insufficient data -Uncertainty -Merging information  Applied Applications -Multitarget observation -Box pushing -Multirobot soccer  Techniques -Reinforcement -Parameter tuning -Particle swarm optimization

32.  Architectures  Communication  Variance  Task Allocation  Learning  Applications

33. o Foraging & Coverage o Flocking & Formations o Box Pushing & Cooperative Manipulation o Multitarget Observation o Traffic Control & Multirobot Path Planning o Soccer

34. Foraging & Coverage

35. Flocking & Formations http://static.guim.co.uk/Guardian/environment/gallery/2007/nov/ 07/wildlife/PD8402359@A-large-flock-of-star-5333.jpg Formation Control

36. Box Pushing & Cooperative Manipulation Kube

37. Multitarget Observation Spletzer and Taylor

38. Traffic Control & Multirobot Path Planning Bruce and Veloso

39. Soccer Bruce and Veloso RoboCup