Quantitative Analysis of Multi-agent Formations: Theory and Applications

Zhu, Guangwei Ph.D., Purdue University, May 2013. Quantitative Analysis of MultiAgent Formations: Theory and Applications. Major Professor: Jianghai Hu. In this thesis, formations of multi-agent systems, including multi-vehicle systems and sensor networks, are studied from a quantitative perspective. First, matrix and scalar valued quantities related to multi-agent system formations are proposed. These quantities are useful in characterizing the robustness, collaborative performance and manipulability of multi-agent systems, are intuitively associated with physical analogies of formations. Their practical value in the analysis of formation control and localization systems is also demonstrated through application examples. With these quantitative characterizations of formations, several optimization problems can be formulated, which are helpful in designing efficient formations for improving the performance of multi-agent systems in various aspects. Finally, based on the concept of stiffness matrix, distributed algorithms are designed to localize agents in the network using angle-of-arrival information. The stability and performance analysis as well as optimization techniques regarding such algorithms are studied using the quantitative framework developed above.