Research On Gait Planning And Control Of Humanoid Robot For Stair Climbing

Humanoid robot has wide applications and strong research value because of its humanoid characteristics.It is necessary for humanoid robot to walk stably in different environments.At present,the gait planning method of humanoid robot is mostly focused on flat ground environment,and directly adopts model-based method to obtain off-line gait.The walking gait designed in off-line state can meet the requirements of smooth and stable walking,but when the robot is on the uneven ground,such as climbing the slope and up and down the stairs,the method can not achieve a good stability control effect.It is easy to cause the robot to fall over because of its unstable posture.Aiming at the gait planning and optimization method of humanoid robot in building climbing environment,it mainly includes generating off-line gait and optimizing gait online with DQN algorithm.The main work is as follows:Firstly,the present situation of humanoid robot research at home and abroad and the existing gait planning methods are analyzed,and the main contents and structure of this paper are introduced.Secondly,the robot leg coordinate system based on D-H matrix rule is established,the forward kinematics equation and inverse kinematics solution are derived,and the two methods for solving inverse kinematics are compared and analyzed.Thirdly,the robot leg model is simplified into a seven-link model,the mobile inverted pendulum model is combined with the variable length inverted pendulum model,and the moving variable length inverted pendulum model based on ZMP is established.The off-line gait parameters of each joint are obtained.Furthermore,combining the gait planning method based on model and intelligent algorithm,a gait stability controller based on DQN algorithm is proposed,and the off-line gait generated by model is optimized on line.After repeated training,the optimal gait data are obtained.At the same time,the stability and energy efficiency are taken as the optimization objectives,and the comparative analysis before and after optimization is obtained.Finally,the virtual simulation system of humanoid robot's lower limbs is built by using the dynamic simulation software Pro/E,ADAMS and the MATLAB/Simulink simulation software,and the proposed algorithm is verified.At the same time,the final optimized gait results are tested on the platform of physical prototype.The experimental results show that the theoretical analysis can achieve a good control effect and improve the stability of the walking process.