Distributed , Decentralized Optimization for Cooperative Control of Multi-Agent Swarm-like Systems

Scientific Systems Co. Inc. and University of California at Berkeley (UCB) propose to develop, test and evaluate a computationally Distributed,Decentralized Optimization for Cooperative Control of Multi-Agent Swarm-like Systems (DOCMASS)such as groups of autonomous Unmanned Air-Vehicles (UAVs). A primary goal of the proposed work is to develop computational models amenable to high-speed simulation, design and control of multi-agent swarm-like systems, and to determine how relatively simple and controllable, interaction rules between individual agents (e.g. UAVs) can lead to sophisticated aggregate behavior which is required; for example, to accomplish autonomous reconnaissance military missions in a hostile terrain. In order to develop, test and implement the DOCMASS system, the following tasks will be carried out: (i) Develop computationally distributed and decentralized cooperative control algorithms for multi-agent swarm-like systems such as groups of UAVs operating under environmental, dynamic, kinematic and communication constraints, (ii) Perform stability analysis of multiple-agent swarm-like systems and develop performance metrics to analyze the effects of inter-vehicle communication constraints on the overall system performance, (iii) Implement and evaluate proposed algorithms on the UC Berkeley multi-agent swarm-like systems simulator, which will be modified to include for a realistic model of an UAV dynamics, sensors and a mobile military network, (iv) Simulation testing on a scenario involving up to two hundred UAVs operating in dynamic and uncertain environment while executing cooperative reconnaissance missions.