Wall-Following Navigation for Mobile Robot Based on Random Forest and Genetic Algorithm

Abstract

Due to the high real-time requirements of robot navigation tasks, the nonlinearity of the robot system itself makes it difficult to model accurately, in addition, the existing “black box” navigation control model is difficult to verify, hard to achieve static and lack of interpretability. Therefore, in this paper, a navigation method for wall-following mobile robots based on random forest and genetic algorithm is proposed. In our proposed method, robots based on perceiving environment information need to autonomously decide the actions to be performed to implement subsequent navigation behaviors. On the UCI robot navigation dataset, the random forest model is first trained, and then control rules are extracted from the trained random forest, and genetic algorithms are used to optimize and filter these rules. Therefore, we have obtained a rule base with good generalization performance, that is, the controller. A comparative experiment on the same data set shows that the method proposed in this paper has a high accuracy rate of 89.19% and a recall rate of 89%. In 6 sets of comparative experiments, compared with the gravity search and particle swarm optimization assisted artificial neural network method and the gravity search and feedforward neural network method, the average accuracy of our proposed method is improved by 16.43% and 29.66%, respectively.