Distributed coordination of multi-agent systems based on estimation over ad-hoc communication networks

A common feature of distributed coordinated multiple agent systems is that multi-agent cooperation is accomplished through interactions between agents wherein information is exchanged across a wireless communication network. Distributed coordinated control of multi-agent systems poses several meaningful theoretical and practical challenges. In this work, we consider three important and related issues: communication logics, swarm cohesion under consensus and the convergence rate of consensus filtering under network throughput limitations.;Optimal communication logics are investigated in Chapter 2 for scheduling the information exchange required to minimize the total energy consumed in the system. We study two different communication logics: an optimal open-loop logic in which each agent periodically transmits information to its neighbors, and a closed-loop logic that schedules transmissions when the local state estimation error is above a preset threshold. We theoretically analyze the optimal deterministic threshold-based logic's performance and compare it with the proposed periodic logic.;Cohesion of multi-agent swarms moving under the control of a consensus filter is studied in Chapter 3. The main result shows that swarming under consensus is cohesive. We establish specific bounds on the degree of cohesion and consensus level. We prove that if the communication graph is regular, then the introduction of integral action into the consensus filter achieves perfect consensus regardless of the number of members.;The convergence rate of consensus filters under throughput limitations is examined in Chapter 4. We present two consensus filter schemes, synchronous and asynchronous. Synchronous consensus obeys the principle wherein individual agents regulate their states only after receiving all neighbors' messages. While, in asynchronous consensus, each agent updates its state when it receives any message from any neighbor. We present a theoretical approach for analyzing the impact throughput limitations have on the convergence rate of the two consensus schemes. We demonstrate the specific advantages of these schemes.;In order to test these coordinated control algorithms in a real multi-robot system, we developed a software multi-robot simulator and real robot testbeds. Our developed simulator and testbeds are introduced in Chapter 5.