Research On Several Problems Of Coordination Control In Multi-agent Systems

A lot of interesting phenomenon of coordinated movement in biological groups can be seen in our daily life,such as a group of pigeons flying above the city,a flock of geese often fly in a V-shaped formation or a line-shaped pattern,the fish in the sea will always keep tight and large groups.In the absence of global information and the collective control,the individuals complete a difficult and complex task only through information exchange between each other.Inspired by the cooperation movement of biological groups,scientists have introduced this mechanism of emerging global behavior by local communication into the area of practical engineering and proposed the cooperative control problem of multi-agent systems.However,when the distance between agents is greater than the sensing radius or agents encounter with some obstacles and complex geographical environment,they cannot communicate with each other,and thus the entire agent network may not be connected.The most of flocking results critically rely on an assumption of connectivity of the underlying network or maintains the multi-agent network by artificial potential function,However,it is not easy or rather impractical to keep connectivity of network all the time.In this thesis,we proposed a dynamic pinning control strategy,when the network is not connected,the whole network is divided into several connected subnetworks at each topology switching time,and then selects a node from each connected subnetworks as the informed agent.In this way,each agent can directly or indirectly receive the feedback from the leader at each moment.Finally,the network is connected as well as the coordination control of multi-agent systems is achieved.Based on the dynamic pinning control strategy,the flocking and consensus control of the multi-agent systems are mainly studied in this work,the main contents of this thesis are as follows:(1)Some basic concepts related to the coordination control of the multi-agent systems,the historical background and the significance of the research are introduced.The recent developments of coordination control of the multi-agent systems are reviewed,in the meantime,some basic theoretical knowledge about the stability analysis of the flocking control and consensus algorithm are prepared.(2)The flocking control of multi-agent systems with switching topology is investigated and a dynamic pinning control algorithm(DPCA)is developed to generate a stable flocking motion for all the agents without the assumption of connectivity or maintains connectivity of network by unbounded potential function.Based on LaSalle Invariance Principle,it is proved that the proposed DPCA ensures that the velocities of all the agents approach to that of the virtual leader,average position of informed agents asymptotically approaches the position of the virtual leader,no collision happens between the agents,and the system approaches to a configuration that minimises the global potentials.In addition,the situation that the agent is dominated by general second-order linear dynamics is considered and a new protocol is proposed to solve the flocking control of multi-agent systems with general second-order linear dynamics.(3)The flocking of multi-agent systems with special conditions that there is no virtual leader and the multi-agent network has switching topology is considered and a pseudo-leader following flocking algorithm is proposed to achieve flocking control of multi-agent system with general second-order linear dynamics.The pseudo-leader is first introduced which is the agent with the lowest fitness in the whole multi-agent network,then the lowest evaluation value agent in each connected subnetwork is selected as the informed agent who can receive the information of the pseudo-leader.Based on the LaSalle invariant principle,it can be proved that the proposed algorithm can achieve the flocking control of multi-agent system.(4)Multi-agent network encounters some obstacles and complex geographical environment in the process of flocking is considered.Some virtual particles are added on the boundary of obstacles or complex geographical environment and then produce repulsive force in the opposite direction with the movement of the agent.Based on the obstacle agent,another new virtual agent is introduced to provide a tangential force and thus allowing the agent to bypass the obstacle more efficiently.The situation of multi-agent network through complex environment is also studied and verifies the effectiveness of the proposed algorithm.(5)The leader-follower consensus problem of multi-agent system with switching topology is investigated and each agent is dominated by general linear dynamics.A dynamic pinning consensus protocol is proposed to achieve consensus of multi-agent systems.The analysis of algorithm is provided with Lyapunov function which shows that the consensus will be achieved asymptotically without assuming that the multi-agent network is connected all the time.In addition,the consensus algorithm is then extended to solve the formation control problem of multi-agent systems.The formation control algorithm not only makes all agents maintain a specific formation but also makes the multi-agent network reach some predefined destination with the same velocity.