Abstractions, Analysis Techniques, and Synthesis of Scalable Control Strategies for Robot Swarms

ions, Analysis Techniques, and Synthesis of Scalable Control Strategies for Robot Swarms Spring Berman Department of Computer Science Harvard University February 28, 2011 Abstract Tasks that require parallelism, redundancy, and adaptation to dynamic, possibly hazardous environments can potentially be performed very efficiently and robustly by a swarm of hundreds to thousands of autonomous robots.Key challenges in developing such a system include the accurate prediction of group behavior and the design of robot controllers that are scalable with the population and provably produce a desired macroscopic outcome. I present a comprehensive approach to modeling a robot swarm, analyzing its behavior, and synthesizing scalable, decentralized stochastic control policies with theoretical guarantees on performance.I demonstrate the application of this approach to the design of a task allocation strategy that does not rely on inter-robot communication, a reconfigurable manufacturing system, and a commercial pollination scenario with micro air vehicles. I also present a study of the mechanics of group food retrieval in ants and a model of this behavior as a strategy for multi-robot collective transport.Tasks that require parallelism, redundancy, and adaptation to dynamic, possibly hazardous environments can potentially be performed very efficiently and robustly by a swarm of hundreds to thousands of autonomous robots.Key challenges in developing such a system include the accurate prediction of group behavior and the design of robot controllers that are scalable with the population and provably produce a desired macroscopic outcome. I present a comprehensive approach to modeling a robot swarm, analyzing its behavior, and synthesizing scalable, decentralized stochastic control policies with theoretical guarantees on performance.I demonstrate the application of this approach to the design of a task allocation strategy that does not rely on inter-robot communication, a reconfigurable manufacturing system, and a commercial pollination scenario with micro air vehicles. I also present a study of the mechanics of group food retrieval in ants and a model of this behavior as a strategy for multi-robot collective transport.