1. A Platform for Local Interactions between Robots in Large Formations Ross Mead Jerry B. Weinberg Jeffrey R. Croxell

2. A Platform for Local Interactions between Robots in Large Formations Motivation Space Solar Power (SSP) Space Solar Power (SSP)  large solar reflector panel in space diameter: ~16.5 km (~10mi) diameter: ~16.5 km (~10mi)  focus concentrated beam of solar energy diverted to an energy plant on Earth for harvesting diverted to an energy plant on Earth for harvesting

3. A Platform for Local Interactions between Robots in Large Formations Motivation One solution that received considerable attention was the use of robots to form a solar reflector. One solution that received considerable attention was the use of robots to form a solar reflector. Imagine the space shuttle releasing thousands of robots, each with a piece of reflector attached to them. Imagine the space shuttle releasing thousands of robots, each with a piece of reflector attached to them. These robots then navigate themselves to form a large parabolic structure resembling a reflector, which is then used to harvest solar energy. These robots then navigate themselves to form a large parabolic structure resembling a reflector, which is then used to harvest solar energy.

4. A Platform for Local Interactions between Robots in Large Formations Motivation  Problem How can a massive collection of robots… How can a massive collection of robots…  ~33,000 required for SSP (Landis 2004) … moving with no group organization… … moving with no group organization…  swarm … coordinate to form a global structure? … coordinate to form a global structure?  formation

5. A Platform for Local Interactions between Robots in Large Formations Problem swarmformation

6. A Platform for Local Interactions between Robots in Large Formations Background Fredslund & Mataric 2002 Fredslund & Mataric 2002 Balch & Arkin 1998 Balch & Arkin 1998 Reynolds 1987 Reynolds 1987 Farritor & Goddard 2004 Farritor & Goddard 2004

7. A Platform for Local Interactions between Robots in Large Formations Background & Goals In related work on formations, units know… In related work on formations, units know…  where they belong in the formation  who their neighbors are supposed to be Goals: Goals:  generality – conforming to a variety of formations  stability – maintaining the formation  robustness – responding to changes in group size  dynamic switching capability – responding to commands for changes in its organization

8. A Platform for Local Interactions between Robots in Large Formations Background This approach to the autonomous control of creating and maintaining multi-robot formations is similar to work done in coordinating formations of Earth-bound, mobile robots. This approach to the autonomous control of creating and maintaining multi-robot formations is similar to work done in coordinating formations of Earth-bound, mobile robots.  Fredslund & Mataric 2002  Balch & Arkin 1998 This work has been inspired by biological or organizational systems, such as geese flying in formation. This work has been inspired by biological or organizational systems, such as geese flying in formation.

9. A Platform for Local Interactions between Robots in Large Formations Background A variety of work has also been done to apply reactive control structures to create emergent group behaviors. A variety of work has also been done to apply reactive control structures to create emergent group behaviors. Flocking algorithms have been used for both physical and simulated robots. Flocking algorithms have been used for both physical and simulated robots.  Ando, et al 1995 A digital hormone model, inspired by biological cell interaction, has also been proposed for robotic organization A digital hormone model, inspired by biological cell interaction, has also been proposed for robotic organization  Shen, et al 2004

10. A Platform for Local Interactions between Robots in Large Formations Background Robot formations have been applied to applications such as automated traffic cones. Robot formations have been applied to applications such as automated traffic cones.  Farritor & Goddard 2004 Swarm behavior control has been applied to urban search-and-rescue robotics. Swarm behavior control has been applied to urban search-and-rescue robotics.  Tejada, et al 2003

11. A Platform for Local Interactions between Robots in Large Formations Formation Control Utilize reactive robot control strategies Utilize reactive robot control strategies  closely couple sensor input to actions Treat the formation as a cellular automaton Treat the formation as a cellular automaton  lattice of computational units (cells)  each cell is in one of a given set of states governed by a set of rules governed by a set of rules

12. A Platform for Local Interactions between Robots in Large Formations Formation Control A command that indicates the geometric formation is sent to a seed robot A command that indicates the geometric formation is sent to a seed robot The formation then transforms as robots… The formation then transforms as robots…  react to changes in their neighbors  attain their calculated relationships based on the formation definition based on the formation definition F ← y = ax 2 seed F ← y = 0 seed

13. A Platform for Local Interactions between Robots in Large Formations Formation Control c ← (x i, y i ) r 2 ← (x-c x ) 2 + (y-c y ) 2 rr F ← y = ax 2

14. A Platform for Local Interactions between Robots in Large Formations A desired formation, F, is defined as a geometric description… A desired formation, F, is defined as a geometric description…  i.e., mathematical function  F ← y = ax 2, where a is some constant Formation Control F ← y = ax 2

15. A Platform for Local Interactions between Robots in Large Formations A robot is chosen as the seed, or starting point, of the formation. A robot is chosen as the seed, or starting point, of the formation. Formation Control F ← y = ax 2 seed

16. A Platform for Local Interactions between Robots in Large Formations Formation Control The desired location on the formation is determined by calculating a relationship vector from c,… The desired location on the formation is determined by calculating a relationship vector from c,…  where c is the formation-relative position (x i, y i ) of the robot, … and the intersection of the function F and a circle centered at c with radius r, where r is the distance to maintain between neighbors in the formation. … and the intersection of the function F and a circle centered at c with radius r, where r is the distance to maintain between neighbors in the formation. c ← (x i, y i ) r 2 ← (x-c x ) 2 + (y-c y ) 2 F ← y = ax 2 rr seed

17. A Platform for Local Interactions between Robots in Large Formations Relationships and states are communicated locally to robots in the seed’s neighborhood, which propagates changes in each robot’s neighborhood in succession. Relationships and states are communicated locally to robots in the seed’s neighborhood, which propagates changes in each robot’s neighborhood in succession. Using sensor readings, robots attempt to acquire and maintain the calculated relationship with their neighbors. Using sensor readings, robots attempt to acquire and maintain the calculated relationship with their neighbors. Formation Control c ← (x i, y i ) r 2 ← (x-c x ) 2 + (y-c y ) 2 F ← y = ax 2 rr seed

18. A Platform for Local Interactions between Robots in Large Formations c ← (x i, y i ) r 2 ← (x-c x ) 2 + (y-c y ) 2 Despite only local communication, the calculated relationships between neighbors results in the overall organization of the desired global structure. Despite only local communication, the calculated relationships between neighbors results in the overall organization of the desired global structure. Formation Control F ← y = ax 2 seed

19. A Platform for Local Interactions between Robots in Large Formations Thus, it follows that a movement command sent to a single robot would cause a chain reaction in neighboring robots, which then change states accordingly, resulting in a global transformation. Thus, it follows that a movement command sent to a single robot would cause a chain reaction in neighboring robots, which then change states accordingly, resulting in a global transformation. Formation Control seed

20. A Platform for Local Interactions between Robots in Large Formations Formation Control

21. A Platform for Local Interactions between Robots in Large Formations Formation Control Likewise, to change a formation, a seed robot is simply given the new geometric description, and the process is repeated. Likewise, to change a formation, a seed robot is simply given the new geometric description, and the process is repeated. F ← y = 0 seed

22. A Platform for Local Interactions between Robots in Large Formations Results A proof-of-concept of the formation control algorithm was successfully demonstrated in a simulated environment at AAAI-06. A proof-of-concept of the formation control algorithm was successfully demonstrated in a simulated environment at AAAI-06.simulated environmentsimulated environment We have developed a robot platform to assess the algorithm in the physical world. We have developed a robot platform to assess the algorithm in the physical world.

23. A Platform for Local Interactions between Robots in Large Formations Robot Platform Each robot features: Each robot features:  a Scooterbot II base differential steering system differential steering system  an XBC v2 microcontroller executes formation control algorithm executes formation control algorithm  a color-coding system and color camera visual identification and tracking of neighbors visual identification and tracking of neighbors  an XBee radio communication module sharing information within a robot’s neighborhood sharing information within a robot’s neighborhood

24. A Platform for Local Interactions between Robots in Large Formations Robot Platform Scooterbot II base Scooterbot II base  precision cut double-decker base rigid expanded PVC rigid expanded PVC strong, but very light strong, but very light  2" risers for additional decks  differential steering system  http://www.budgetrobotics.com/ http://www.budgetrobotics.com/

25. A Platform for Local Interactions between Robots in Large Formations Robot Platform Differential steering Differential steering  two modified R/C servo motors with 2 1/2" diameter rubber wheels if motors are operated at same speed, the robot goes straight if motors are operated at same speed, the robot goes straight if motors are operated at different speeds, the robot turns or spins if motors are operated at different speeds, the robot turns or spins r r - w / 2 r + w / 2 w

26. A Platform for Local Interactions between Robots in Large Formations Robot Platform XBC v2 microcontroller XBC v2 microcontroller  executes formation algorithm  back-EMF PID motor control  fast charging ~1 hour to fully charge ~1 hour to fully charge  http://www.botball.org/ http://www.botball.org/

27. A Platform for Local Interactions between Robots in Large Formations Robot Platform Color-coding system Color-coding system  visual identification and tracking of neighbors Color camera Color camera  multi-color, multi-blob simultaneous color tracking Start y Stop y ID y Start y - ID y ID y - Stop y Start y Stop y Start y - Stop y Robot ID = ID max * (Start y - ID y ) / (Start y - Stop y )

28. A Platform for Local Interactions between Robots in Large Formations Robot Platform Color camera Color camera  multi-color, multi-blob simultaneous color tracking Color-coding system Color-coding system  visual identification and tracking of neighbors Start y Stop y ID y Start y - ID y ID y - Stop y Start y Stop y Start y - Stop y Robot ID = ID max * (Start y - ID y ) / (Start y - Stop y )

29. A Platform for Local Interactions between Robots in Large Formations Robot Platform XBee radio communication module XBee radio communication module  sharing state information within a robot’s neighborhood  ZigBee/IEEE 802.15.4 specification  up to 65,535 nodes on a network  support for multiple network topologies  low duty cycle  long battery life  collision avoidance  retries and acknowledgements  link quality indication  128-bit AES encryption  http://www.maxstream.net/ http://www.maxstream.net/

30. A Platform for Local Interactions between Robots in Large Formations Robot Platform XBee radio communication module XBee radio communication module  share information within a robot’s neighborhood  ZigBee/IEEE 802.15.4 specification  300’ (100m) line-of-sight range  peer-to-peer, point-to-point, point-to-multipoint and mesh network topologies  retries & acknowledgements for error handling  65,535 network addresses for each channel  http://www.maxstream.net/ http://www.maxstream.net/

31. A Platform for Local Interactions between Robots in Large Formations Robot Platform Simple, light, and inexpensive… Simple, light, and inexpensive…  reproduction of each unit is easy and affordable A successful implementation of the algorithm on a modest number of physical robots will prove that the approach is viable in the real world. A successful implementation of the algorithm on a modest number of physical robots will prove that the approach is viable in the real world.

32. A Platform for Local Interactions between Robots in Large Formations Future Work – Formation Management Develop a graphical user interface to provide a human operator with… Develop a graphical user interface to provide a human operator with…  a visualization of the formation  information on each individual robot unit

33. A Platform for Local Interactions between Robots in Large Formations Future Work – Dynamic Neighborhoods Implement an auction-based method to determine neighborhoods dynamically… Implement an auction-based method to determine neighborhoods dynamically…  a robot is chosen to be a neighbor based on its distance to the desired location in the formation

34. A Platform for Local Interactions between Robots in Large Formations rr F ← y = x√3F ← y = -x√3 F ← y = 0 seed Future Work – Formation Classification Classify different types of formations… Classify different types of formations…  those defined by multiple functions  those that generate erroneous neighbors

35. A Platform for Local Interactions between Robots in Large Formations Future Work – Formation Classification

36. A Platform for Local Interactions between Robots in Large Formations Future Work – Formation Classification

37. A Platform for Local Interactions between Robots in Large Formations Future Work – Formation Classification

38. A Platform for Local Interactions between Robots in Large Formations References Balch, T. & Arkin R. 1998. “Behavior- based Formation Control for Multi- robot Teams” IEEE Transactions on Robotics and Automation, 14(6), pp. 926-939. Balch, T. & Arkin R. 1998. “Behavior- based Formation Control for Multi- robot Teams” IEEE Transactions on Robotics and Automation, 14(6), pp. 926-939. Bekey G., Bekey, I., Criswell D., Friedman G., Greenwood D., Miller D., & Will P. 2000. “Final Report of the NSF-NASA Workshop on Autonomous Construction and Manufacturing for Space Electrical Power Systems”, 4-7 April, Arlington, Virginia. Bekey G., Bekey, I., Criswell D., Friedman G., Greenwood D., Miller D., & Will P. 2000. “Final Report of the NSF-NASA Workshop on Autonomous Construction and Manufacturing for Space Electrical Power Systems”, 4-7 April, Arlington, Virginia. Farritor, S.M., & Goddard, S. 2004. “Intelligent Highway Safety Markers”, IEEE Intelligent Systems, 19(6), pp. 8- 11. Farritor, S.M., & Goddard, S. 2004. “Intelligent Highway Safety Markers”, IEEE Intelligent Systems, 19(6), pp. 8- 11. Fredslund J., & Mataric, M.J. 2002. “Robots in Formation Using Local Information”, The 7th International Conference on Intelligent Autonomous Systems, Marina del Rey, California. Reynolds, C.W. 1987. “Flocks, Herds, and Schools: A Distributed Behavioral Model, in Computer Graphics”, 21(4) SIGGRAPH ’87 Conference Proceedings, pages 25-34. Tejada S., Cristina A., Goodwyne P., Normand E., O’Hara R., & Tarapore, S. 2003. “Virtual Synergy: A Human- Robot Interface for Urban Search and Rescue”. In the Proceedings of the AAAI 2003 Robot Competition, Acapulco, Mexico.

39. Questions? For more information, visit the exhibition or http://roboti.cs.siue.edu/projects/formations/