Research on Automatic Navigation System Construction Based on SLAM Algorithm and Deep Neural Network

Abstract

For autonomous mobile robots, it is the prerequisite to be able to locate and navigate autonomously in various complex environments. By analyzing the shortcomings of traditional multilayer perceptron and back propagation algorithm. A brand-new network structure based on SLAM algorithm and deep neural network is designed, which mainly realizes the modulus and fixation of weights, so as to accelerate the training speed. By using SLAM technology and deep neural network to limit the error accumulation of inertial navigation system, the performance of integrated navigation system can be significantly improved and the incremental map of environment can be built online. Simulation results show that this method makes full use of the information provided by simultaneous mapping and positioning, and effectively improves the accuracy of navigation system.