Walking Balance Control Of Humanoid Robots

Humanoid robot has legged walking mechanism similar to that of human,giving it the potential to travel freely through all kinds of complex human environments.Therefore humanoid robot has received widespread attention from the robot research community.However,humanoid robot is still uncomparable to human in terms of locomotion ability,especially in terms of walking stability,which cannot meet the requirements of practical applications.The walking stability of humanoid robot is influenced by many factors,including both the robot hardware performance factors and the external environment factors.In this thesis,the walking balance control algorithms against the nonideal characteristics in the actuating mechanisms and against the uneven floor in the environment are proposed.The proposed methods can significantly improve the walking stability and the robustness against the uneven terrain under the same hardware conditions,contributing to improving the walking performance and to expanding the application scenarios of the humanoid robot.The main contents of this thesis includes:1)Modeling of the nonideal characteristics in the humanoid robot's actuating mechanisms.A set of modeling methods for the nonideal characteristics in the actuating mechanisms are proposed,which can obtain accurate nonideal characteristics models in the absence of direct measurements.2)Deformation compensations for the humanoid robot during the walking process.A set of feedforward compensation methods are proposed,which can effectively compensate the deformations caused by the nonideal characteristics in the actuating mechanisms,in accordance with the expected force bearing condition of the feet and the hips during the walking process.3)Vibration suppression for the humanoid robot during the walking process.A vibration suppression method based on input shaper and another method based on optimal control are proposed,which can both effectively reduce the vibration of the center of mass around its reference position during the walking process.4)Humanoid robot walking on uneven terrain.A set of methods for humanoid robot walking on uneven terrain are proposed,which enable the robot to achieve stable walking on stochastic uneven terrain without laser or vision sensor information.The effectiveness of all the proposed methods is validated through walking experiments on the humanoid robot Kong-I.The last chapter concludes the main contents of this thesis,and addresses the plan for the future work.