Particle Robotics

Natural and biological systems inspire novel approaches to robotic design and control. This thesis applies principles of stochastic mechanics and collective intelligence to develop amorphous robots composed of loosely coupled components, or particles. Like the individual units that constitute many biological structures or swarms, the particles lack a unique identity or specialized function, and they operate without a centralized control. Only through interactions and external conditions do complex behaviors arise. To provide greater scalability and robustness, individual particles are kept simple, capable of a single degree-of-freedom motion that can be modulated; alone they are incapable of directed locomotion. However, by loosely coupling and systematically modulating the particles, the aggregate can migrate as a single entity and adaptively reconfigure when interacting with unfamiliar environments. We call this stochastic formation a particle robot. The particle communication and coordination does not rely on the unique identity or addressable position of individual particles, thereby removing any single point of failure typical of traditional robots. Further, groups of particles may splinter into smaller groups or annex additional particles without catastrophic effects. Through detailed modeling of the interactions and dynamics of the particles and extensive simulations based upon this modeling, the work presented in this thesis characterizes the scalability, robustness, resilience, and adaptability of this paradigm.