Developing a Low-Cost Robot Colony Felix Duvallet, James Kong, Eugene Marinelli, Kevin Woo, Austin Buchan, Brian Coltin, Christopher Mar, Bradford Neuman 20 mins pres + 5 mins questions = 25 minutes AAAI Fall Symposium 2007 November 10, 2007 Fix CMU logo – it looks out of place right now (probably move RC logo up or down)
Outline Motivation and Goals Robots Sensors Infrastructure Behaviors Sustainability Future work
Motivation and Goals Motivations – A low-cost experimental platform for multi-agent research – Tool for testing AI algorithms Need capable robots – Open-source software to promote similar research elsewhere Goal: Develop a colony of robots – Low-cost robots – Homogeneous architecture – Distributed algorithms Motivation and Goals Two hard things: experimental, and low-cost Remove open-source software?
Colony Robot BOMDragonfly Board Motors ORBs Range- Finders XBee Module Charging Contacts Robots Cheap (~$350) but capable
Colony Brains ATMega 128 – 8MHz max – 128Kbytes program memory Programmed in C – arv-libc, avr-gcc – open-source, multi- platform tools Robots
Sensors Sharp IR rangefinders BOM Also used: – Bump sensors – Pyroelectric sensors Sensors
BOM – Bearing and Orientation Module IR emitter/detector ring Emitter mode – All emitters are powered simultaneously (beacon) Detector mode – Detectors can be polled individually for analog intensity readings – Most excited sensor indicates bearing to emitting robot Co-planarity assumption Must coordinate BOMs Build connectivity graphs 1 20 Sensors
Wireless Infrastructure Integrate BOM and wireless network – Beaconing robot sends wireless message – Receiving robots know when to poll BOM Token-ring coordination scheme Share information about neighbors – Propagate global connectivity graph Infrastructure
Swarming Local obstacle avoidance Low coordination Highly scalable Behaviors
Lemmings Simple follow-the- leader Medium Coordination Variant: – Cooperative maze solving Behaviors
Towards Self-Sustainability Goal is to develop a self-sustainable robot colony that can operate unassisted for long periods of time Requirements – Autonomous recharging – Remote monitoring and control Sustainability
Autonomous Recharging Robots must request charging resources from a charging station Additional sensor aids homing sequence Daughter-board controls battery charging process Video? Sustainability
Bay Allocation Maximize distance between occupied bays – Minimize likelihood of collisions Can you really spend a whole slide explaining this? – perhaps if you make a very nice diagram Sustainability
Probably remove slide – diagram isn’t useful with time constraint
ColoNet Towards Remote Monitoring Autonomous recharging makes extended-duration tasks possible Need tools to: – Monitor progress – Identify and resolve problems Added bonus – Useful debugging tool User Server TCP/IP Server Wireless R0R0 R1R1 RnRn Sustainability Web client
Future Work Increase task complexity – Coordinated manipulation – Environment exploration Increase trial duration – Days or weeks Increase Colony scale I don’t think you should put SLAM here, it’s very (very x 100) far away from working, and you don’t want to make promises you can’t keep Future Work
Colony Members Duncan Alexander Ben Berkowitz Austin Buchan Brian Coltin Felix Duvallet Siyuan Feng Aaron Johnson Jason Knichel James Kong Christopher Mar Eugene Marinelli Brad Neuman Suresh Nidhiry Justin Scheiner Greg Tress Kevin Woo Cornell Wright Prof. George Kantor
Questions? Can we get an updated picture like this one?