Local Path Planning of Mobile Robot Based on Long Short-Term Memory Neural Network
Na Guo, Caihong Li, Di Wang, Yong Song, Guoming Liu, Tengteng Gao
Abstract
In order to solve the local path planning problem of mobile robot in the unknown environment, more complex control algorithms and solutions are needed for various special situations; however, they cannot be suitable for every environment. A local path planning algorithm of the mobile robot based on Long Short-Term Memory (LSTM) neural network is proposed to simplify the complexity of control, increase the ability of learning and generalization, and improve the efficiency of path planning. Firstly, the model structure of LSTM neural network is designed based on the task requirements of the local path planning of mobile robot. Then, the training data needed by the neural network are collected based on the fuzzy control algorithm and used to train the LSTM model. Finally, the trained model is tested according to the local path planning task. The experimental results show that compared with the fuzzy control algorithm, the designed method improves the calculation velocity, and can adapt to the environment with more obstacles compared with the BP neural network algorithm.