Research On Multi-robot Target Searching Based On Intelligent Optimization Algorithms In Unknown Environments

With the development of robot application,the research on cooperative search targets of multi-robot systems has been widely utilized in many applications,such as resource exploration,military reconnaissance,anti-terrorism and mine clearance,post-disaster search and rescue,etc.At present,the requirement of intelligent control for mobile robots is getting more and more urgent,and the self-organizing mechanism of intelligent optimization algorithms can be adaptively matched with multi-robot systems.The application of intelligent optimization algorithms to multi-robot collaborative search has significant advantages.In this paper,to solve the multi-robot targets searching problems in unknow environments(the obstacles and targets position are not the prior information),a series of intelligent optimization algorithms are thoroughtly studied.Specific research work and contributions are summarized as follows:(1)To solve the static single target searching problems of multi-robot in unknown environments,a novel multi-swarm hybrid method based on particle swarm optimization(PSO)and Fruit Fly optimization algorithm(FOA),named multi-swarm hybrid FOA-PSO algorithm(MFPSO),is proposed.Firstly,an adaptive inertia weight is introduced to improve PSO algorithm.Then,an improved fruit fly optimization algorithm with multi-swarm adaptive coefficients is proposed,which provides a better selection mechanism for the improved PSO to find the next optimal robot position.After that,the multi-swarm strategy is introduced to enlarge the search scope,improve the diversity of robot search,and effectively avoid premature convergence and falling into local optima.In addition,when the robot falls into the local optima,the multi-scale cooperative mutation escape mechanism can improve the escape ability and obstacle avoidance ability of the robot,in this way,the search speed and success rate of the robot can be effectively improved.From the experimental results,it can be seen that the performance of MFPSO is better than the comparative methods(Adaptive Robotic PSO,A-RPSO;Robotic PSO,RPSO),especially in multi-swarm robots and large-scale environments.(2)To solve the static single target searching problems of multi-robot in unknown environments,an improved bat algorithm for multi-robot target searching,named adaptive robotic bat algorithm(ARBA),is proposed.Firstly,the adaptive inertia weight strategy adopted by ARBA helps to improve the diversity of robot search and provides an effective mechanism for escaping from local optima.Then,the Doppler effect is introduced to improve the frequency formula so that the robot can get adaptive compensation when it moves,which helps robots avoid premature convergence.Moreover,since the location of the target in the environment is unknown,multi-swarm strategy introduced into the ARBA can further improve the diversity of robot search and expand the search scope,which allows the robot to find the target faster than the existing algorithms.As a control mechanism of robots,ARBA helps robots avoid obstacles and complete multi-robot target searching task in unknown environments.The experimental results verify that the ARBA is obviously superior to other comparative methods(Robotic Bat Algorithm,RBA;A-RPSO)and shows better performance.The proposed method has the advantages of less iterations,high success rate,high efficiency,and smoother searching trajectories.(3)To solve the static and dynamic single-target searching problems of multi-robot in unknown environments,an improved grey wolf optimizer(GWO)for multi-robot target searching,named adaptive robotic GWO algorithm(RGWO),is proposed.Firstly,the best learning strategy is introduced to improve the position updating formula of GWO,so that the algorithm can be suitable to the actual mobile situation of the robots,and the searching robots can move towards the target(prey)step by step.Then,RGWO adopts adaptive inertia weight.By increasing the “aggregation degree” or decreasing the “evolution speed”,the influence of inertia weight will increase,which is helpful to maintain the diversity of the robot search and solve the premature convergence problem.In addition,due to the escape of the prey,the pursuit robot is more likely to fall into local optima.In order to avoid the search robots from falling into local optima due to the escape of the target,the RGWO adopts an adaptive speed adjustment strategy and introduces an escape mechanism.The RGWO is verified and compared with the previous single-target searching method(MFPSO,ARBA).The MFPSO and the ARBA have good and similar performance in the number of iterations,success rate and efficiency.In contrast,the RGWO has obvious advantages over the previous single-target searching method(MFPSO,ARBA)in terms of success rate,the number of iterations and efficiency,and can better complete static and dynamic target searching task.But the searching trajectories of RGWO is not smoother than MFPSO and ARBA.(4)To solve the problem of multi-target dynamic task assignment and cooperative search for multi-robot systems in unknown environments,an improved self-organizing map(SOM)neural network method combining dynamic window algorithm(DWA)is proposed.Firstly,the SOM algorithm is improved.A locking mechanism is established on each competing neuron to make up the defects of its self-organizing behavior,so as to avoid the robot oscillation and hovering.Then,to solve the problems of obstacle avoidance and speed jump,the competing neurons weights of SOM are updated by using adaptive DWA.With this algorithm,the obstacle avoidance is realized without knowing the obstacle position information,and the robot's motion trajectory is optimized.In addition,when the location of the target and obstacle changes,the method can reallocate the task and re-plan the searching path in real time,and complete the cooperative searching of dynamic multi-target in the unknown environments.The simulation results show that the method has less iterations,higher success rate and efficiency,and can better complete multi-target dynamic task assignment and cooperative searching,showing better robustness.(5)When conducting the multi-robot searching task in unknow environments,the robots need to obtain environmental map information through sensors.To solve this issue,a multi-robot grid map merging method based on speed up robust features(SURF)is proposed.In this method,the robot motion coordinate system is transformed into a rigid body.The mathematical model of the grid map merging problem is represented by the minimization problem of image registration and the mathematical model is built.Firstly,an improved SURF algorithm is used to extract features from the grid maps.Secondly,the random sampling consensus(RANSAC)algorithm is employed to eliminate the mismatched data and get the initial merging parameters,and the merging parameters are used as the initial value of the iterative nearest point(ICP)algorithm to solve the objective function.The experimental results show that the proposed method can achieve reliable merging of grid maps established by multi-robot and the proposed method has good robustness,high precision and fast merging speed.This method can be applied to multi-robot target searching,which can further improve the performance of searching algorithm.