Research On Synchronized Flocking Control And Pinning Control Of Mobile Agent Network

Inspired by biological aggregating phenomenon in nature, some emergent mechanismson swarming behavior have been discovered and used to design multi-agent system inengineering field. With the development of multi-agent system, more and more researchersmake an attempt to study global characteristics of scalable multi-agent system in the view ofcomplex network and the object network composed of large-scale mobile agents is called asmobile agent network. In the field of mobile agent network, researchers discuss thecorrelation between dynamic characteristic of swarming behavior and features of networktopology, and the theoretical results can be used to design optimized flocking control law forlarge-scale engineering swarm system, such as wireless sensor network, space probes andunmanned aerial vehicles system.This dissertation focuses on synchronized flocking and pinning control of mobile agentsnetwork, and the emphases will be optimization of synchronization performance of dynamicflocking behavior while improving connectivity and invulnerability of global networktopology and reducing the power consumption of mobile agents during the evolutionaryprocess. Main contributions are structured as follows:Firstly, basic concepts and research tools on mobile agent network are introduced; somepopular models and algorithms on flocking control are summed up; the analysis method ofsynchronization, the measure of synchronizability and Lyapunov stability theorem fordynamic flocking behavior are given. All above are theoretical basis of follow-up study.Secondly, the relationship between sensing radius of agents and invulnerability measureof network topology is discussed. Inspired by regulating the sensing radius to coordinate thepower consumption with the performance of network, we propose a cost function includingedge information of mobile agent network as evaluation objective and use PSO algorithm tooptimize it to get a best sensing radius for mobile agents. The simulation results show thefeasibility and effectiveness.Thirdly, considering mechanical characteristic of nonholonomonic agents in actualapplication, a path planning strategy based on Bezier Curve Method is proposed to satisfy theconstraints of guide angle and curvature. Combined with Minimal Circumcircle Method toensure connectivity of an initially connected network during evolutionary process, severaloptimal flocking strategies with lower consumption or better synchronizability for multiplenonholonomic robots are proposed. The simulation results show the feasibility andeffectiveness.Finally, the pinning control is introduced into mobile agent network. The agents movetoward the target state followed by the informed agents which get information from the virtual leader. This paper proposes an optimal pinning control strategy which the informed agents arechoosen by the largest degree principle and the position of informed agents can be regulateddynamically to maintain the control power, then the network can be pinned to the target stateflexible. The simulation results show the feasibility and effectiveness.