AbstractsEngineering

In this thesis project, we consider the cooperative control of multi-agent systems under limited communication between the individual agents. In particular, quantized values of the relatives states between neighboring agents are used as the control parameters for each agent. As an introductory part, the theoretical framework for the distributed consensus problem under perfect communication is reviewed with the focus on the system stability and convergence. We start from the common problem setup that single integrator agents with a static tree communication topology, where the stability constraints and convergence guaranty are derived for dierent quantization models: uniform, logarithmic and dynamic. Then the conclusions are extended to switching tree topologies, tree topologies with disconnected time intervals and nally general undirected graphs. The control performance like  onvergence rate and the area of convergence set are compared between systems with and without quantization eects, and also among the systems with dierent quantizers. Furthermore, similar techniques are applied to other system dynamics with quantized control inputs. We investigate additional constraints on the stability of the corresponding discrete time system due to the presence of quantization eects. Explicit upper bounds on the sampling time that guarantee convergence are derived. As expected, the sampling frequency has to be increased accordingly under dierent quantization models. The multi-agent system composed of second-order agents under general undirected communication graphs is also taken into account with quite dierent analytical tools. Finally we switch to an alternative model that takes the relative quantized states as control parameters instead. Distinctive convergence properties are found between these two models and detailed comparisons are made. Throughout this report, all results obtained are supported by numerical simulations.