COORDINATION OF MULTI-AGENT SYSTEM IN CONSTRAINED SPACE

Abstract

This thesis is to investigate coordination of multi-agent system in constrained space. It consists of three main parts: environment exploration, environment perception and formation control. A task allocation strategy is designed to explore an unknown environment and topology mapping. Two approaches of environment perception for obstacle avoidance are proposed, including the introduction of Dirac delta function to design the potential field of spatial constraints, and development of a speech control system. Formation control is studied via the virtual-leader-based and the leader-follower-based approach. In the first structure, a model-based and a NN-based controller are derived, respectively. A Formation Potential Field is designed to automatically generate the optimal formation. In the second structure, a vision-based solution is provided under visibility constraints. An event-triggered coordination mechanism is designed considering communication limitations. A formation scaling factor is designed to scale up/down the formation size when the region is impassable with the original formation size.