The Research And Implementation Of Walking Planning And Control For Humanoid Robots

Humanoid robots have always been a popular research direction in the field of intelligent robotics due to their unique appearance and structure as well as flexible and powerful functions.The realization of the humanoid robots' walking function also provides the basis for the development of other upper-level functions.Although there exist a series of theories and methods for the stable walking of humanoid robots in flat ground environments,the performance of walking pattern generation on non-flat ground still needs to be improved.Aiming at the problems,this paper studies walking planning and control for humanoid robots,and effectively plans reasonable footings in uneven terrain environments.Meanwhile,combined with deep reinforcement learning methods,the traditional preview control theory is improved and a new preview controller is designed,which allows humanoid robots to walk stably using the planned path.The main researches of this paper include the following aspects:1)Proposing adaptive walking path planning algorithm based on different terrain environments.Sensors are used to perceive the environment and generate point cloud maps of grounds from which parametrized descriptions are computed.According to different ground environment,a variety of gait sets are designed for adaptively searching paths,and path planning on uneven surfaces is efficiently completed.2)Proposing walking pattern generation based on deep reinforcement learning and preview control.First,this paper analyzes the shortcomings of traditional preview control theory.Meanwhile,an improved preview control algorithm based on deep reinforcement learning is designed to solve the problem that it is challenging for humanoid robots to walk stably in uneven environments.In addition,according to the walking characteristics of the humanoid robot,the arm swinging strategy is proposed to help to make the walking of humanoid robots more stable and natural.3)Designing an improved walking controller for humanoid robots.Based on the proposed preview control algorithm,a new walking controller combined with the actual gait planning and balance control method is brought up to plan and control the walking of humanoid robots on uneven terrains.4)Implementing a walking control system for humanoid robots,including map creation module,path planning module and walking motion module,which realizes the operational interaction between the user and the humanoid robot.The experimental results show that the path planning method proposed in this paper can effectively search for a reasonable footing for robots in uneven or even complex environments.The method based on deep reinforcement learning and preview control proposed in this paper can relatively successfully conduct planning and control for humanoid robots on uneven terrains.