Research On Autonomous Obstacle Avoidance Algorithm For Humanoid Robot In Indoor Environment

Recently,autonomous obstacle avoidance has become a hot issue of the research on mobile robot field.The autonomous obstacle avoidance ability of the robot can relieve the limitation of the space,so that it can no longer serve the human in the fixed area and further expand its application scope.With the ability of robots has been improved and indoor environment has become more complex,robots need to perceive the environment and develop mobile strategies.Therefore autonomous obstacle avoidance is an important ability of mobile robot.In this paper,the humanoid robot NAO is taken as the research object,and the FastSLAM algorithm is used to realize simultaneous localization and mapping,and the autonomous obstacle avoidance strategy is proposed under the static environment and dynamic environment respectively.The main research content is as follows:Firstly,the global and local coordinate systems is built,and the humanoid robot motion model and observation model is established based on coordinate system.The principle of FstSLAM algorithm is introduced and the effect of different number of particles on the performance of the algorithm is verified by simulation.Secondly,a static obstacle avoidance strategy based on fuzzy Q-learning is proposed.The Q table is initialized based on fuzzy inference system,which make the robot learning on the basis of daily experience.Thus the fuzzy Q-learning has a faster convergence rate.And it is applied to the autonomous obstacle avoidance of the NAO robot in an unknown environment with obstacles.In order to overcome the negative effect of the robot motion control noise,the fractional order PI controller is used to correct the deviation.The simulation results show that the algorithm is efficient and feasible.Finally,to realize the obstacle avoidance of humanoid robot in the dynamic environment with moving obstacles,this paper proposes a dynamic obstacle avoidance strategy based on variable dimension flower pollination algorithm.The grid map is used to describe the external environment,and the shear type map updating strategy is adopted to improve the update efficiency.The fitness function is designed to evaluate the planning path reasonably which combining the advantages of Chebyshev distance and Euclidean distance.The performance and feasibility of the proposed algorithm are verified by simulation and experiment.The research work in this paper can provide theoretical basis and method for the development of mobile robot,which has important theoretical value and scientific significance.