Local Path Planning Of Mobile Robot Based On Intelligent Control

Autonomous navigation technology is one of the core and focus of robot applications,mainly including mapping,positioning,perception and path planning,among which path planning is the key technology in mobile robot navigation.Local path planning usually has the problems of local deadlock and path redundancy due to unknown obstacle information and limited detection environment of sensors configured by robots,which make it difficult to quickly plan a safe and smooth shortest path.Based on a variety of intelligent control methods,this paper studies how the mobile robot completes the local path planning task efficiently and autonomously.Aiming at the problem of local path planning task and improving the efficiency of path planning for mobile robot,the local path planning method based on fuzzy control and Long Short-Term Memory(LSTM)neural network is studied.Firstly,a local path planning algorithm of mobile robot based on fuzzy logic control(FL)is proposed for the task of local path planning.A fuzzy controller is designed,which combines the common obstacle avoidance behavior and walking along the wall behavior.According to the problem of local deadlock and path redundancy,the solution strategies of cumulative turn sum,trap prediction mechanism and Finite State Machine based on artificial potential field method are designed.Then,in order to improve the efficiency of path planning,a local path planning algorithm of mobile robot based on LSTM neural network(LSTM?FL)is proposed.The LSTM neural network model is designed,and the training data is collected through the FL algorithm to train the model.Finally,the two methods are simulated and tested,respectively.The simulation results show that the two methods can safely avoid obstacles and reach the target in a variety of common obstacle environments.Compared with FL,LSTM?FL has a significant improvement in planning efficiency.In view of the limitations of the sensing environment of the sensors configured on mobile robots,this paper proposes a fusion method of local path planning for mobile robots based on LSTM neural network and reinforcement learning(LSTM?FTR)to avoid the unreachable target,shorten the path distance and enhance the ability of learning and generalization.Firstly,a neural network model including LSTM units is designed for the local path planning task of mobile robot.Secondly,the training data is collected by FL algorithm with low dimensional input,and the network model is pre-trained by transfer learning to learn the basic ability to solve the local path planning problems.Then,combined with reinforcement learning,the model can learn new rules from the environment to better adapt to different scenarios.Finally,the final model is simulated and tested.Compared with FL and LSTM?FL algorithm,LSTM?FTR improves the success rate of path planning,shortens the length of the path,and better reflects its advantages in dense and complex environments.