Research On Path Planning Method Of Indoor Mobile Robot

With the continuous development and innovation of science and technology,intelligent mobile robots using artificial intelligence are quietly integrating and changing the life of human life.The aging of the population and the hot demand of smart homes have gradually increased the research investment of home-based mobile robots,and the requirements for path planning technology have gradually increased,which has determined the urgency and necessity of path planning technology in the field of robot research.On the basis of previous research work,this thesis makes some further research on the path planning problem of indoor mobile robots.By comparing the advantages and disadvantages of path planning algorithms with different characteristics,it is concluded that a single algorithm cannot complete path planning in both static and dynamic environments.It is finally determined that we use the combination of RRT algorithm and Q-Learning algorithm for path planning,and improved methods of global path planning and local path planning are proposed respectively.The main research work in this thesis is as follows:In terms of global path planning,traditional RRT algorithms have problems such as poor optimality,large randomness and slow convergence rate,which affects the overall path planning results.Combining the advantages of artificial potential field method and RRT algorithm,this thesis proposes an RRT path planning algorithm based on artificial potential field method,which introduces the idea of gravity based on target point in artificial potential field method to RRT algorithm,and continue to introduce adjustable step size and two-way search growth ideas.This improve search efficiency and planning efficiency on the basis of improving obstacle avoidance ability.Finally,the global path planning of mobile robot based on improved RRT algorithm is realized in simulation experiment and physics experiment.In terms of local path planning,the standard Q-Learning algorithm based on Markov decision process in reinforcement learning lacks prior knowledge of the environment,which leads to the problem of slow training speed and low iteration efficiency in the learning process.This thesis proposes a Q-Learning path planning algorithm based on artificial potential energy field,which uses the artificial potential energy value to initialize the Q value to solve the blind search problem in the early stage of learning.At the same time,through the initialization in the early stage of learning,multiple "trial-and-error" tests on obstacles can be avoided,and finally better path planning effects can be obtained with fewer training times,thus indicating that the improved algorithm have faster convergence and better optimization in the learning process.Finally,the local path planning of mobile robot based on improved Q-Learning algorithm is realized in simulation experiment and physics experiment.Finally,in order to combine these two path planning algorithms and apply the path planning theory to practice,an experimental platform for indoor mobile robot path planning system is built,and the optimal path planned by the improved algorithm is verified by simulation and physics.The experimental verification proves that the simulation results in the ideal state are compared with the verification results of the indoor mobile robot experimental platform in the actual environment,and the feasibility of the relevant theoretical algorithm is verified.The experimental results further verify that the indoor mobile robot is in the static environment and the dynamic environment.The path planning can be completed safely and efficiently.