Statistical Robotics: Controlling Multi Robot Systems using Statistical-physics

In a multi-robot system consisting of numerous agents, it may be impractical to individually identify each agent. Consequently, issuing specific commands to each agent might not be feasible. We therefore, introduce the concept of a Statistical multi-robot system (SMRS). Such systems comprises a very large number of agents that cannot be identified or located individually. Moreover, Since it is impractical to track and communicate the complete configuration of an SMRS, we resort to statistical physics methods, specifically gas kinetic knowledge, to extract their distribution. But unlike in Thermodynamics, we employ the fact robotic agents can sense their environment, communicate their microscopic state, and change their local behavior to enable control. The concept of an SMRS suggests that the comprising agents should be as simple as possible, for practical reasons. In this study, we demonstrate how an SMRS comprised of single-degree-of-freedom agents can be controlled by a global controller. Using the same rationale, we define a successful mission of an SMRS as one in which a sufficient portion of the agents accomplish the mission. To demonstrate the efficacy of our approach, we provide a motion planner and exemplify our formalism in both simulations and real-world experiments. One-Sentence Summary: To control huge multi robotic systems we resort to the theory of statistical-physics and utilize its macroscopic properties.