Research On Flocking Control Of Multi-agent Network Via Pinning

Since classical flocking model of swarm system is firstly proposed by Reynolds, expertsfrom areas respectively survey the cause of flocking phenomenon and its possible engineeringapplication. Recently, the idea of pinning control has been applied to flocking control andobtained significant achievements with the development of pinning control of complexdynamic network. This dissertation mainly focuses on flocking control of multi-agent networkvia pinning on the foundation of the summary for the related research result. The main workand contributions are as follows:1、In existing flocking control algorithm of multi-agent network via pinning, the networkmay appear "splitting" phenomenon in process of evolution. Therefore, a flocking controlalgorithm of multi-agent network via local optimization and pinning is proposed. Thisalgorithm is presented by joining local optimization algorithm in flocking control algorithm ofmulti-agent network via pinning, considering there are only part of the agents (calledinformation agents) have the information of the virtual leader, they continuously move towardthe direction that has the largest algebraic connectivity of network consisting of all theneighborhood of them in current moment and next moment. It improves the connectivity ofthe whole network through local optimization. Finally, better performance is obtained by thecomparison with the previous methods. Theory analysis and numerical experiment furthershow that the proposed algorithm can effectively keep global connectivity of multi-agentnetwork during evolution. All the agents gather around the virtual leader in smaller area and itcan reflect relative importance of the information agent in multi-agent network.2、Then a flocking control algorithm for power-driven multi-agent networks via localestimation and pinning is further proposed based on the research of the complex power-drivencontrol system, as the existing flocking control via pinning strategies have shortcomings inpreserving topology connectivity and decreasing power consumption. Firstly, a poweriterative consensus estimation algorithm is used to dynamically estimate the algebraicconnectivity of multi-agent network. Secondly, the algorithm is dynamically adjustedindividual launch radius according to the estimation of algebraic connectivity and the powerrestriction. So that it can preserve global connectivity of the multi-agent network duringmotion and effectively reduce the power consumption. Finally, the simulation results provethe effectiveness of the proposed control scheme.3、A flocking control algorithm for multi-agent network via community division andpinning is proposed through the research of multi-agent network with randomly initializationdistribution on the effect of flocking control via pinning, considering initial distribution of network consists of several communities, which has obvious community structure. Firstly, themulti-agent network is divided into some communities based on hierarchical clustering and anode importance evaluation algorithm is presented to select the most important node of eachcommunity as the information agent. Then a control algorithm is presented by preservingcommunity connectivity to achieve the whole multi-agent network keeping connectivity inprocess of evolution. It has proved that the algorithm improves the effectiveness of flockingcontrol via pinning through specifically choose pinning nodes.