| dc.description.abstract | Biology has brought much enlightenment to the development of human technology, for example, the collective behaviors inspired engineering applications (such as, the unmanned vehicle formation, the satellite alignment etc.), and even the study of network theory. This discipline has made a significant contribution to technology development. As a prospective solution to the current issues, multi-agent control has become a popular research topic in recent decades. The traditional control methods based on the classical models are suffering from high sensitivity to model accuracy, computational complexity, low fault tolerance, and weakness in real-time performance. Therefore, the advantages of multi-agent control are obvious: 1) easy maintenance and expansion of the system by repairing, replacing or adding agents; 2) high fault tolerance and robustness, ability to function properly even when some agents fail; 3) low requirement of distributed controllers, which brings low cost and large flexibility.
In this thesis, I investigate problems on modeling and control of multi-agent systems. In particular, I propose a three-dimensional model to simulate collective behavior under high-speed conditions. I design an improved adaptive-velocity strategy and weighted strategy to enhance the performance of the multi-agent system. Moreover, I analyze the performance from the aspects of energy and parameter space. I show how the model works and its advantages compared to existing models.
Then, I study the design of distributed controllers for multi-agent systems. Output regulation with input saturation and nonlinear flocking problems are studied with the assumption of a heterogeneous switching topology. The output regulation problem is solved via low gain state feedback and its validity verified by theoretical study. Then, the flocking problem with heterogeneous nonlinear dynamics is solved. A connectivity-preserving algorithm and potential function are designed to ensure the controllability of the multi-agent system through the dynamic process.
Overall, this thesis provides examples of how to analyze and manipulate multi-agent systems. It offers promising solutions to solve physical multi-agent modeling and control problems and provides ideas for bio-inspired engineering and artificial intelligent control for multi-agent systems. | en |
