Humanoid Robot Autonomous Walking Based On SLAM In Indoor Environment

With the promotion of computer technology, robotics technology and intelligencetheory, the extensive research of intelligent robots are carried out internationally.Humanoid robot is closed to the original meaning of robot and its research has beenwidespread concerned. Humanoid robot is one of the most active researches in roboticareas. Humanoid robot is focused on the complex environment for some advanced humantasks in the current research, such as human assistance, family care and transportation.Therefore, the research of humanoid robotic capability of autonomous navigation has veryimportant significance.Based on the humanoid robot NAO as the research object, the simultaneouslocalization and mapping(SLAM), the deviation and obstacles problems in the process ofwalking are discussed. In this paper, EKF-SLAM algorithm, fractional order controlmethod and Q-learning algorithm are used for humanoid robot autonomous walking, andthe simulation and experiment have been carried based on humanoid robot NAO. Specificresearch contents are as follows:Firstly, for the simultaneous localization and mapping problems of NAO, theEKF-SLAM algorithm based on laser sensor is adopted to realize the autonomous walkingand Mahalanobis Distance probability method is proposed to solve the data association ofSLAM. In order to solve the deviation problems produced by interference in the process ofwalking, the fractional order PI controller is designed for closed loop compensation and toreduce mobile deviation. Simulation research shows that the fractional order PI controllercan effectively reduce the deviation.Secondly, because NAO can encounter the obstacle during walking, Q-learningalgorithm is used for path planning. By establishing checkerboard path planning modeland choosing an action based on the current state, the next moment state and return valuecan be assessed. So the best avoidance path is determined. The simulation results showthat the Q-learning algorithm effectively solves the NAO robot obstacle avoidanceproblem in the process of autonomous walking.Finally, the indoor environment is selected as the external environment of NAO robot walking in this article, and the experiment is divided into two groups:(1) the indoorenvironment contains landmarks;(2) the indoor environment contains landmark andobstacle. By the experimental verification, NAO robot not only can walk autonomouslyand finish the simultaneous localization and mapping, but also avoid the obstacle.