Formation Control And Collaborative Searching Of Swarm Robotics

Swarm robotics is a particular approach to multi-robot systems.It originates from biomimetics of natural social creatures and has broad application prospects in many areas,including manufacturing,rescue,ocean exploring,military defense and so forth.Generally,the swarm members are in low-complexity,which limits their capacities of communication and perception.How to reach emergent collaborative behaviors under those constraints is a crucial problem.In particular,collaborative searching is a fundamental research aspect of swarm robotics.It belongs to the category of coordinate moving and decision making,i.e.,the whole swarm needs to move cooperatively and finally reach the consensus to complete the task.Eefficient distributed control and cooperation methods are essential to solve the coordinate moving and decision making problems.This dissertation focus on the research of collaborative searching methods for swarm robotics under the constraints of relative localization,limited perception,and local communication,etc.Concretely,the searching process will be considered as two stages:preliminary searching and refinement searching.According to the swarm biomimetic and corresponding constraints,the architecture of the system will be given,and the problems will be stated and solved,i.e.the formation control and distributed processing problems for the preliminary searching stage and constrained metaheuristic searching problem for the refinement searching stage.The proposed methods will be evaluated qualitatively and quantitatively through theoretical derivations,simulations and entity verifications.Start with the characteristics of swarm robotics,the mechanisms of relative localization,communication,and coordinated control will be analyzed.Moreover,a four-layer architecture for swarm robotics will be presented.Also,a two-stage model which includes preliminary searching and refinement searching will be given based on corresponding hypnosises and constraints.Then the problem statements of formation control and distributed processing for the preliminary searching stage,and the particle swarm optimization(PSO)based metaheuristic method for the refinement searching stage will be introduced.Secondary,based on the review of the state-of-the-art approaches to the formation control problem,a cascade leader-follower formation control strategy will be proposed.This method considers the limits of the field of view,local communication as well as member kinematics.It can form the line and V-shaped formations in a cascade way and have the potential to apply to large scale swarms.The simulation results will be given for evaluating the effectiveness,robustness,flexibility of the proposed method.Furthermore,by considering the space-time property of the line and V-shaped moving formation,a distributed processing paradigm based on a modified 2-D Roesser model is proposed.This method enables global signal processing with just simple communication between the nearest neighbors.As an example,a visual salient region detection algorithm will be applied to this processing paradigm to achieve collaborative searching with the moving formations in the preliminary searching stage.The performance of the introduced method in several situations will be evaluated by the simulation experiments.Finally,for the refinement searching stage,an extended particle swarm optimization(PSO)based method will be proposed for the metaheuristic approach.By introducing the limits of the field of view,relative localization,local communication,and kinematics,a constrained adaptive robotic PSO method is proposed.This method will follow the concept of evolution speed and aggregation degree from adaptive PSO algorithm.However,the aggregation degree will be determined by a united degree which not only related to the fitness values but also the robot numbers in a time-varying characteristic swarm.Furthermore,the initial weight will be calculated by combining the new aggregation degree and evolution speed.Different to the direct update of particle positions in standard PSO,the proposed method will update the forward speed and turn speed as control inputs of swarm members under a non-holonomic model of mobile robots.All the above methods will be tested on the E-Puck mobile robot platform.In general,this dissertation will study the collaborative searching problem of swarm robots with the combination of theories of swarm robotics modeling,formation control,multi-dimensional system,visual saliency,and particle swarm optimization,etc.Theoretical derivation and experimental results will be given.It will shed some lights on the theoretical and practical aspects of the collaborative searching problem of swarm robotics.