Collective Plume Tracing: A Minimal Information Approach to Collective Control

Abstract

A team of simple robots are used to trace a chemical plume to its source in order to find buried landmines. The goal of this paper is to analyze and design "emergent" behaviors to enable the team of simple robots to perform plume tracing with the assistance of information theory. The first step in the design process is to define a fundamental tradeoff for collective systems between processing, memory, and communications for each robot in order to execute the desired collective behaviors. The baseline problem is to determine the minimum values for processing, memory, and communications simultaneously. This will enable the designer to determine the required information flow and how effectively the information resources are being utilized in collective systems. The solution to this baseline problem is an 8-bit processor, no memory, and three words to communicate. The details of this solution are described in this paper. The second step is to extend the previous kinematic solution to a more general Hamiltonian based solution to develop a Fisher Information Equivalency. The Fisher Information Equivalency leads directly to necessary and sufficient conditions for stability and control of nonlinear collective systems via physical and information exergy flows. In particular, the creation of a "virtual/information potential" by the team of robots via the decentralized distributed networked sensors produces a direct relationship between proportional feedback control, stored exergy in the system, and Fisher Information. This nonlinear stability formulation through Fisher Information directly provides an optimization problem to "tune" the performance of the team of robots as a function of required information resources.