Coordination of distributed autonomous systems
Jie Lin, A. Stephen Morse · 2004
This thesis focuses on the analysis of emergent behaviors of a multi-agent system. The ultimate aim is to synthesize complex group behaviors from simple social inter-actions among individuals based on simple strategies. One of the many challenges in the study of emergent behaviors of multi-agent systems is characterizing the local properties of each agent that will lead to global emergent behaviors. Two types of emergent behaviors are described, namely, the Multi-agent Flocking Problem and the Multi-agent Rendezvous Problem. In both problems, provably correct distributed local rules are proposed. In the Flocking problem, a local heading average rule was studied and convergence of the group heading to a common unspecified heading was proved using results in convergence of matrix products in Markov Chains. Replacing the word agent by a more general notion of identity naturally leads the Flocking problem to a class of network consensus problems. In the Rendezvous problem, convergence of the group positions to a common unspecified location was established using combinations of convex analysis and standard Lyapunov function approach. An asynchronous version of the Rendezvous problem was also solved by modelling the process using a suitably defined hybrid model and a procedure called Analytic Synchronization. Recent applications of the Controlled Mobility concept to wireless network in both synchronous and asynchronous settings to optimize energy consump-tion is briefly discussed at the end.