Distributed formation for large swarm of agents
Date of Issue2013
School of Electrical and Electronic Engineering
Recent advances in sensor technology have led to the development of complex networked control systems consisting of large number of interacting agents, thus opening up new opportunities and challenges in analysis and control of such complex systems. Recently, formation control of a large swarm of networked multi-agent system has received considerable attention. Due to the large size of the system and the communication limitations of the complex networks, traditional formation control methods fail to demonstrate satisfactory performance. This thesis presents several methodologies to solve distributed formation control problem for a large swarm of networked multi-agent system. Formation control of network of agents with single-integrator dynamic can be simplified as a consensus problem. This thesis firstly addresses the distributed consensusability problem in complex networked agent systems based on the structural observability of network systems. A notion of unobservable node is introduced which leads to structural unobservability, and the relationship between the existence of unobservable nodes and the distributed consensusability of network systems is explored. Subsequently, a simple tool is presented to identify the unobservable nodes of a network system which is equivalent to examining the distributed consensusability of the network system. Moreover, a simple mathematical method is proposed to modify a distributed consensusable network into a weight-balanced one. To obtain formation control of large-scale multi-agent systems with general Euler-Lagrange dynamic equations, the shape formation technique has been presented. However, in existing shape formation control problem, having access to the desired reference of the shape by all agents is a requirement. Due to limitations in communication bandwidth and range when operating in a large group, the agents are not able to have access to the desired reference through broadcasting. To overcome these problems, an estimator is presented to construct the desired reference for each agent based on local information exchange by its neighboring agents. In addition, two distributed shape formation control methodology is proposed to deal with bidirectional and unidirectional interactions among agents. The existing methods on shape formation control are limited to simple formation problems by single group of agents. The main obstacles that prevent the achievement of the complex shape formation are difficulties in representing complex formations. This problem is overcome by introducing a multi-group coordination methodology. The multi-group coordination problem is formulated into two parts: inter-group formation where groups form a desired formation and intra-group formation such that members of each group form a specified shape. To guarantee the stability of multi-group system, adaptive interactive forces are proposed to cope with inter-group interactions. Finally, a distributed dynamic multi-group methodology is presented for complex shape formation control of a large swarm of agents. To obtain the distributed dynamic multi-group formation, a global virtual leader is assigned to perform the overall planning and local virtual leaders are assigned to obtain their desired information from the global virtual leader in a distributed way. Therefore, the problem is formulated into two parts: global group behavior and local group behavior. The proposed method is effective in achieving distributed complex formations, distributed multi-group reconfiguration, and distributed time-varying shape formations for large swarm of agents.
DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering