Research On Path Planning Technology For Mobile Robot Based On Sampling-Based Algorithm And Genetic Algorithm

Mobile robots are one of the most active fields in scientific and technological innovation,which is a comprehensive research result of information processing,sensor fusion,automatic control,and artificial intelligence.With the deepening of robot technology research,the performance and function of mobile robots are also improving.Mobile robots are widely used in industries such as medical care,agriculture,and service,but also have played a positive role in dealing with complex environmental operations,natural disasters,space exploration,national defense and other fields.Therefore,the research of mobile robot technology not only has good economic value but also has great strategic significance.As one of the key technologies in the field of robotics,path planning is not only the foundation of autonomous navigation of mobile robots but also the foundation of improving the autonomy and intelligence of robot movement and operation.Among the existing path planning algorithms,the outstanding performance of sampling algorithms in high-dimension path planning problems and genetic algorithm in multiobjective optimization problems has attracted much attention in the research circle.Although there are many research achievements on these two kinds of algorithms,there is still much room for improvement in convergence speed.Based on sampling algorithms and genetic algorithm,this paper conducts an in-depth study on the path planning of mobile robots.The main research contents are summarized as follows.(1)Aiming at the problems of long time-consuming for RRT* algorithm to obtain the initial path and low efficiency of sampling method,an MRRT-RRT* path planning algorithm based on RRT*-Smart algorithm and Informed RRT* algorithm is proposed.Firstly,by improving the initial path acquisition methods of the RRT*-Smart algorithm and the Informed RRT* algorithm,an initial path acquisition method of running the RRT algorithm several times and selecting the shortest path is proposed.The proposed method overcomes the slow convergence of the RRT*algorithm.Secondly,combining with the sampling advantages of RRT*-Smart algorithm and Informed RRT* algorithm,a fusion sampling strategy is proposed,in which the neighborhood of the current shortest path is sampled at a fixed frequency,and the rest is sampled in the ellipsoid with the starting point and end point as the focus and the length of the current shortest path as the major axis.The sampling range is continuously reduced to improve the sampling efficiency of the algorithm.Then,the probabilistic completeness and asymptotic optimality of the proposed MRRT-RRT* algorithm are proved,demonstrating the feasibility of MRRT-RRT* algorithm.Finally,the MRRT-RRT* algorithm is compared with Quick-RRT* algorithm,RRT*-Smart algorithm,Informed RRT* algorithm,and RRT* algorithm.The results show that the proposed MRRT-RRT* algorithm significantly improve the convergence speed and reduces the memory consumption rate.(2)Aiming at the low efficiency of sampling method of RRT* algorithm,combining with the effective guidance of the artificial potential method to the sampling process,a PQ-RRT* path planning algorithm combining P-RRT* algorithm and Quick-RRT* algorithm is proposed.Firstly,the artificial potential field method is integrated into the whole sampling process to provide effective guidance for the sampling process and reduce the number of iterations.Secondly,on the basis of improving the sampling process,the optimization framework of the Quick-RRT*algorithm is adopted to further shorten the cost value of tree nodes.Then,the proposed PQRRT* algorithm is proved to own probability completeness,asymptotic optimality,and fast convergence,and the feasibility of the proposed algorithm is proved theoretically.Finally,the PQ-RRT* algorithm is compared with the P-RRT* algorithm,Quick-RRT* algorithm and Informed RRT* algorithm in four typical environments.The results show that PQ-RRT* algorithm can obtain higher quality initial solutions and has a faster convergence speed,which verifies the effectiveness of the proposed algorithm.(3)Aiming at the problem that the low quality of initial population leads to poor performance of genetic algorithm,a population initialization method based on directed acyclic graphs and effective obstacles is proposed,which own stronger generality and higher efficiency.Firstly,two directed acyclic graphs generated from the starting point and the ending point are constructed,and the information of effective obstacles is innovatively introduced into the construction process of directed acyclic graphs,shortening the construction time of directed acyclic graphs.Secondly,half of the population is generated from the starting point directed acyclic graph,and the other half is generated from the end point directed acyclic graph,which improves the diversity of the initial population.Then,the proposed population initialization method is compared with the existing population initialization methods.The evaluation indexes are the path length of the best individual in the initial population and the execution time of generating the initial population.The results show that the proposed population initialization method has stronger universality and higher quality of the initial population.Further,to verify the initial population affect the performance of the genetic algorithm,the proposed population initialization method of two directed acyclic graphs and the population initialization method of single directed acyclic graph are applied to the genetic algorithm.The results show that the population initialization method of two directed acyclic graphs significantly improve the convergence speed and path quality at convergence.