Cooperative Control of Networked Multi-Agent Systems

Abstract

In recent years, technological advances and the latest trends in communication and control systems engineering have driven the revolution for multi-agent systems with distributed control through consensus algorithms for various applications. Numerous real-time problems and applications are associated with the complex interconnection control of the distributed multi-agent systems. As the scalability of the networks is expanding dramatically so it is important to understand the dynamic behavior of networked systems for continuous availability. These systems require subsystems at the agent level to make autonomous decisions by exchanging information with neighboring agents to achieve their local and global convergence objectives. Simultaneously, this leads to practical and theoretical challenges in multi-agents due to their wireless nature, limited bandwidth, communication delays, random topologies, and limited resources. One of the objectives of this Ph.D. study was to develop an efficient consensus algorithm for distributed cooperative control of the multi-agent systems. Also, the author focus was to find the agents in the network by effective estimation. A fully distributed autonomous framework is considered where agents are restrained with limited communication resources. Distributed consensus algorithms are addressed for fixed and dynamic network topologies that quickly generate arbitrarily accurate estimates. Reliable and unreliable communication was also taken into account in the asynchronous time update in the communication network. Throughout this thesis, matrix and graphic theories are used as prerequisites for the design and analysis of distributed algorithms. Necessary robust convergence conditions are considered to ensure the multi-agent-system dynamics to seek the convergence. Finally, the proposed algorithm is utilized for spectrum sensing in a cognitive radio network. For a particular goal, this research present an idea for energy and network model. The proposed energy model is responsible for spectrum sensing of a cognitive radio network to determine the presence of primary users. As the spectrum is detected free or occupied, secondary users of the network build efficient topologies that reduce both network latency and interruptions for appropriate measurements in reaching consensus. In the end, sufficient conditions for the convergence along with the best possible estimates are computed for the proposed and the existing systems. Also, Improvements and validation of the proposed algorithms in multiple network scenarios are tabulated at the end of each chapter. Furthermore, numerical gain analysis and the simulation results reflect the effectiveness and efficiency of the proposed method based on the key parameter indicators, which includes the total number of iterations, processing time, asymptotic convergence factor, and asymptotic convergence time. It is worth mentioning here that maximum 45% performance gain is achieved by the proposed method while comparing with metropolis method and similarly, maximum 37% performance gain is achieved when comparing with the local degree weights.