Study On Autonomous Mobile Robot Path Planning In Dynamic Environment

As China’s demographic dividend fades, the robot is increasingly in a prominent role in the social life, and our country is paying more and more attention on the robot industries. The study on the path planning of mobile robot navigation is one of the core issues, and the path planning work is a prerequisite for robots to perform tasks. As robot environments become more complex, especially when the robot’s freedom increased, traditional performance of path planning algorithm based on deterministic space description of falling sharply, and even lead to "dimension of disaster". In order to solve this problem, researchers applied random sampling techniques to path planning, proposed algorithms such as Probabilistic Roadmaps Method (PRM), Rapidly-Exploring Random Tree (RRT), and finally achieved good results. However, when the robot workspace was with more obstacles or narrow passages, efficiency would be poor. In response to the problem, researchers have put forward methods like Gaussian PRM, bridge testing PRM, etc., but there are still redundant and poor sampling results of sampling problems left.This paper studies on the PRM path planning based on sampling improvement in order to improve the efficiency. Major work includes the following:1. This research introduces the study of PRM planning algorithms by focusing on basic PRM algorithm theory, and makes a detailed analysis of the efficiency of the Gaussian PRM, obstacle PRM, and bridge-testing PRM.2. This research proposes an incremental sampling strategy aiming at the weakness of the existing method. This proposed method uses small amounts of sampling points on the preliminary sample in the configuration space, constricts the range sampled according to the information, and determines the region need to sample by sampling test. Finally, the sampling efficiency is increased.3. A four-point test method for sampling in two-dimensional planning space is designed. This method uses four-point to identify obstacles in order to avoid redundant collection of the sunken area. Then, the Four-point method is used in progressive sampling PRM, simulation shows that the results are good.4. In order to improve the precision and stability of path tracking. After analyzing the structure of wheeled robot and its path tracking features, and constructing the kinematics model of wheeled mobile robot, a Fuzzy-Neural Network (FNN) based path tracking method with two-stage control is built. In the first stage, a FNN controller determines robot’s turning radius by processing robot’s pose information. In the second stage, the controller adjusts angular and linear velocities by making use of the turning radius and the condition of the path ahead. The experiments show that the controlled robot can track the planned path accurately and robustly when it runs at high speed; the process of path tracking is stable and no slipping and rolling occur.This research simulates the proposed methods in Matlab and OMPL platform. At the end of this paper, the full text is summarized, innovation and key results of the study is described, and the issues that need further study are pointed.