Research On Mobile Robot Path Planning Method In Complex Environment

The path planning technology of mobile robots is a key subject in the navigation technology of autonomous mobile robots,and also an important prerequisite for the application of mobile robots in the real world.Therefore,the study of path planning is necessary.In this paper,the traditional algorithms are studied,followed by the A*(A-Star)algorithm and the RRT*(Rapidly Exploring Random Tree-Star)algorithm.For the special environment where there is a narrow space,a heuristic fusion algorithm based on the two is proposed.After that,the feasibility of the Deep Reinforcement Learning method is analyzed and studied.Finally,the PPO(Proximal Policy Optimization)algorithm is combined with the Intrinsic Curiosity Module(ICM)to conduct path planning research in the Unity 3D simulation environment.The main research work of this article has the following two points:1.From the perspective of traditional algorithms,the characteristics,advantages,and disadvantages of various path planning algorithms are analyzed.The A*algorithm is the most effective and direct method to solve the shortest path,but the calculation amount will increase significantly when facing a large area map;RRT series algorithms have the characteristics of fast,complete probability,and strong adaptability,but when the target point is in a narrow space,the convergence rate decreases.Aiming at this special environment,this paper proposes a heuristic fusion algorithm based on the A*algorithm and the RRT*algorithm.When exploring a narrow space,RRT*first performs a global search,and performs path stitching according to a heuristic function,and then performs a local search through the A*algorithm.Finally,the path points are smoothed based on the gradient descent method.The algorithm has been simulated and verified in MATLAB.The proposed heuristic fusion algorithm combines the accuracy of the A*algorithm and the rapidity of the RRT*algorithm to achieve faster efficiency.2.The feasibility of the algorithm is analyzed through the relevant theoretical foundation of Deep Reinforcement Learning.With reference to the complex actual hazardous chemical environment,a dynamic simulation environment was built in Unity 3D.In this environment,a Deep Reinforcement Learning algorithm that combines the PPO algorithm,and the Intrinsic Curiosity Module is used.The PPO algorithm interacts the agent with the environment and optimizes the neural network with the goal of maximizing rewards.And use multi-threading to speed up the convergence rate of the strategy in the continuous domain.The Intrinsic Curiosity Module trains the forward network and the reverse network to realize the autonomous learning of the agent in a sparse reward environment by predicting the environment's state and agent's actions.After many simulations,it is shown that the algorithm can realize the autonomous learning of path planning strategies by mobile robots in the built environment.