Research On Coordinated Control Of Hybrid Multi-Agent Systems

The hybrid system is an important branch of complex systems.Because of its extensive engineering application background,it has become a hot topic in the research of multi-agent coordination.In this dissertation,algebraic graph theory,matrix theory,control theory and other theoretical tools are used to study the consensus problem and winner-take-all competition problem for hybrid multi-agent systems.The main research results can be summarized as follows:1.Consensus problem for first-order hybrid multi-agent system consisting of continuoustime and discrete-time dynamic agents is investigated.In this system,for each continuoustime dynamic agent,it exchanges information with its neighbors with discrete-time dynamics only at the sampling time,while it can interact with its neighbors who have continuous-time dynamics in real time and obtain its own states in real time? for each discrete-time dynamic agent,it updates its state at the sampling time,and the information interactions between it and its neighbors all happen at the sampling time,too.According to the aforementioned interaction modes among different agents,a suitable consensus protocol is designed for this hybrid multi-agent system.By using mathematical tools such as graph theory and differential mean value theorem of matrix function,a criterion is obtained which can guarantee the hybrid multi-agent system can reach consensus under the designed protocol.This research result establishes a unified framework for the consensus of continuous-time and discrete-time multi-agent systems.2.Consensus problem for second-order hybrid multi-agent system is studied,where the system is composed of continuous-time and discrete-time dynamic agents.According to whether the continuous-time dynamic agents can obtain their own position and velocity values in real time,two effective consensus control protocols are designed for this hybrid multi-agent system,respectively.Based on mathematical tools such as algebraic graph theory,matrix theory and system transformation method,sufficient and necessary criteria are established for solving the consensus problem of second-order hybrid multi-agent system.The unified framework is established in the consensus problem for the discrete-time and the sampled-data second-order multi-agent systems.3.Consensus problems for two kinds of hybrid multi-agent systems with heterogeneous dynamics are studied,respectively.The first kind of hybrid multi-agent system is composed of first-order discrete-time and second-order continuous-time dynamic agents.According to whether the second-order continuous-time dynamic agents can obtain their own position and velocity information in real time,two types of effective protocols are presented for this hybrid multi-agent system,respectively.Then,by using graph theory and system transformation method,the necessary and sufficient criteria are obtained for solving the consensus problem for this hybrid multi-agent system.The second one consists of first-order continuous-time and second-order discrete-time dynamic agents.Based on different interaction modes among different agents,two kinds of effective protocols are also proposed for this kind of hybrid multi-agent.Meanwhile,the necessary and sufficient conditions are obtained for solving the consensus problem of this multi-agent system.4.A class of heterogeneous multi-agent systems is studied,which can show the phenomenon of winner-take-all competition.First-and second-order continuous-time dynamic agents coexist in this multi-agent system.Two different types of protocols with and without velocity measurements are proposed for this multi-agent system,respectively.Firstly,all equilibrium points of the proposed multi-agent system are solved by using the Lasalle’s invariant principle.Then,the Lyapunov indirect method is used to analyze the stability of all equilibrium points.Finally,through analysis,it is known that the agent with the largest initial input will become the only winner and remain active all the time,and the remaining agents will eventually lose activity to zero.The results reveal that winner is independent from the dynamics of agents,but is only determined by initial inputs.