Research On Coordinated Control And Formation Of Multi-agent System

The multi-agent system behaves in comparatively complicated coordination withlocal interactions between different agents. Few study results have been achieved onfinite-time synchronization between nonlinear agents or finite-time consensus of themulti-agent system. Formation control methods lack accuracy and stability in practicalapplications. Path planning algorithms lose globally optical solutions easily in specificformation application. Therefore, the study on multi-agent formations and coordinationcontrol is of theoretical and practical significance.Using time-dependent continuous state feedback protocol, the dissertation solvesthe problem in finite-time consensus of the multi-agent system and finite-timesynchronization between non-linear agents. According to synchronization andconsensus theories, multi-agent system formations can be controlled accurately andstably. Based on accurate and stable formations, artificial potential field algorithm(APFA) for wall-following navigation built on switching topologies and formationchanges is proposed, which has overcome the shortcomings that formation pathplanning algorithm traps into local optimum easily.Finite-time synchronization of nonlinear dynamics systems are solved by use oftime-variant gain continuous state feedback, obtaining a sufficient condition for theissue. By converting the fractional-order nonlinear systems to the equivalentinteger-order models, projective synchronization of the master-slave system in differentfractional-order non-linear dynamic systems has been realized by adopt of Lyapunovstability theory and state error feedbacks. Agents in multi-agent system behave in stateof nonlinear dynamics; therefore synchronization in nonlinear dynamics systems can beregarded as a special case in the multi-agent system consensus. Studies on nonlineardynamics synchronization and finite-time synchronization in the dissertation not onlygives effective solutions to relative issues in integer-order and fractional-order nonlineardynamics systems, but also provides theoretical basis and solutions for followingsections.For finite-time consensus in the multi-agent system, after analyses on directed andundirected graphs, sufficient conditions are obtained on the basis of time-dependentcontinuous state feedback, which supports consensus of the first-order and second-ordersystems in any given limited time. Most approaches on finite-time consensus of multi-agent systems before this dissertation improve the2ndsmallest eigenvalues ofLaplacian matrix by adjusting the communication topology structure, while the methodin this dissertation concerns protocol performance without adjustment, which proves tohave simple arithmetic, convergence time presetting, generalized application and otheradvantages.For multi-agent systems in different lags and switching topology, under conditionof satisfying given H∞performance index, this dissertation proposes a sufficientcondition for first-order multi-agent system consensus by use of Linear MatrixInequality (LMI). For issues in multi-agent formation and coordinated control like noisedisturbance, communication delay and switching topology, methods presented helpsto analyze the effect of external disturbance on system consensus, thus to improve the ofdisturbance resisting capacity.The dissertation solves problems in formation control of the multi-agent system bythe formation diagram method based on synchronization and consensus theories.First-order and second-order system consensus theories ensures stability in formationcontrol, and the design for expected formation structure ensures formation accuracy,which requires information interaction between only few adjacent agents for overallformation consensus. Underwater robots formation and UAVs formation have beenstimulated by Matlab and experimental verification has been conducted in2Dsimulation platform and hardware platform for underwater robots.This dissertation proposes improvements for two obvious defects in formation pathplanning by use of traditional APFA. It presents to update APFA by wall-followingnavigation built on switching topologies and formation changes to prevent traditionalAPFA from being local optimized. An adaptive maneuvering target tracking algorithmbased on the “the current statistics” model is proposes to detect he state ofmaneuvering targets, as improvements for updating potential functions.This method hassolved lack of tracking and detecting in traditional APFA, which normally leads toerrors in potential functions.