Research On First-order Multi-agent Systems With Dynamic Quantizers

The design of control systems becoming increasingly enormous andcomplicated. Nowadays, an equivalent multi-agent system receives more and moreattention because of its convenience in analysis and design. Considering for thecommunication constraints and the limited computing resources in control system,researches on quantizers have become a hot topic. This, consequently, leads theresearches on multi-agent systems with quantizers to be a main trend in moderncontrol systems. In the application of multi-agent systems, most of the tasks arerealized by making the agent states come to consensus. It is reported that theadhibition of quantizers can make the multi-agent systems reach consensus in finitetime, and, naturally, the quantization consensus becomes a priority. This paper aimsat studying the quantization consensus in first-order multi-agent systems.Firstly, some mathematic tools and control theories such as graph theory of thecommunication topology, matrix theory, stability theorems of systems, and thedefinition of quantizers are introduced. At the same time, we introduce multi-agentsystems and some existing control algorithms, and analyze their disadvantagesthrough simulation results. Then, the necessity of investigating the multi-agentsystems with dynamic quantizers is derived.Secondly, an existing consensus strategy is revise by utilizing dynamicquantizers and zero-order holders to produce a hybrid system. Then, event triggeredmechanism for updating dynamic quantizers is introduced to guarantee theconsensus in the corresponding system. Thus, a centralized event-triggeredquantized control strategy for multi-agent systems is proposed. The MATLABsimulation results show the agreement with our analysis, and some guidance forpractical application is given.Finally, a distributed event-triggered quantized control strategy for multi-agentsystems is put forward by taking the actual limitation in practical applicationsettings for centralized class into consideration. That the control input of agent i isonly related to the states of itself and its neighbors makes the investigation inconsensus theory be more close to the practical engineering application. Thesimulation results not only verify the effectiveness of the proposed algorithm, butalso present its rapid convergence property.