Hopping And Adaptive Balance Control Of Bionic Legged Robot Based On Series Elastic Actuators

Legged robot is an important research field in mobile robotics.Compared with traditional wheeled robots,legged robots have both advantages and challenges in executing investigating,rescue and transporting tasks in complicate and unstructured environments.In recent years,the researches of legged robots have made a great progress,but the applications of legged robots into daily life are still restricted by the limitation of the speed of locomotion,the ability of disturbances rejection and the endurance.Focusing on improving the energy efficiency,interactivity and adaptability of legged robots,resorting to the methods and theories of adaptive control,this paper studies the joint control,hopping control,disturbances rejection control and dynamic balance control problems of legged robots.First,compliant joints based on Series Elastic Actuator(SEA)are designed to improve the torque servo performance and energy efficiency of the joints,and a hopping algorithm based on this type of joints is proposed.Then,as the main contribution of this paper,to improve the robustness of the controllers,a series of adaptive control algorithms to control joint,balance and dynamic motion are proposed to imitate the sensing and learning process of human-being in body motion control.The effectiveness of these controllers is validated on various types of SEA based robots,and these works provide a new concept to design and control legged robots.The main contributions of this dissertation are:1.Design a joint based on SEA with its calibration method of SEA joint,and propose two adaptive control algorithms based on EKF observer for the velocity/position control problem with load variations.The effectiveness of the proposed methods is validated via experiments on a single SEA joint.2.To reduce the redundant work of coding different identification program for robot of different structures,propose a versatile recursive dynamics estimator algorithm that can generate different identification model for different types of legged robots.The correctness and effectiveness of the algorithm is validated through both simulations and practical experiments.3.To reduce the loss of accuracy in the virtual model control based biped robot with load variations and unknown disturbances,derive an equivalent identification model for biped robot in standing phase and propose a standing balance controller that combines the virtual model control algorithm and online identification algorithm to guarantee the accuracy of robot posture control without sacrificing the compliance of the robot.The effectiveness of the controller is validated on a double leg robot with SEA joints.4.To reduce influence of weak torque tracking performance of SEA based joints when the robot is hopping,propose a criterion to evaluate the limitation of SEA joints according to angular momentum tracking performance and designed a hopping control algorithm for SEA based legged robots.Experiments are conducted on a single leg and a double leg robot to validate the effectiveness of the algorithm,and consecutive hopping with average CoT of 0.38 is implemented.5.To reduce the bandwidth constraints and delay influence on SEA joints when the robot tries to maintain balance during hoping,propose a learning algorithm for dynamic balance problem to control the angular momentum of torso by planning torque profiles of SEA joints during hopping.An experiment is conducted on a single leg robot to validate the effectiveness of the controller.