Research On Anti-disturbance Control And Trajectory Optimization For Hydraulic Driven Biped Robot Based On Model Predictive And Direct Collocation Method

In the field environment,the legged robot can adapt to the rugged and complex terrain such as mountains and gullies because of its discontinuous contact with the ground,and thus have a larger range of ground activities than the wheeled and tracked mobile robots.Compared with quadruped robots,the leg number of biped robots is relatively small,which not only saves costs and facilitates promotion,but also has a similar motion mode to humans,and can move freely in various complex environments of human activities.And human have been hoping to create a similar to our own robot used for entertainment,education,health care,military,industrial manufacturing,severe hazardous environment and rescue,etc.Therefore,biped robots have huge development potential and have become a research hotspot in the field of robotics,and represent the comprehensive development level of robotics.In recent years,launched by Boston Dynamics of the United States,the hydraulically driven biped robot Atlas has fully demonstrated its superior results in dynamic motion stability,coordination,explosive power,and hardware system integration.However,due to the confidentiality of its algorithm technology,research teams around the world have been inspired to research biped robots with passion.Based on the hydraulically driven biped robot studied by our research group,aiming at the requirements of its high-speed dynamic walking motion stability and anti-disturbance control,according to the ideas and research levels gradually in-depth from the bottom to the top,this paper carried out the joint electric hydraulic actuator control,dynamic control and trajectory online optimization of bipedal robot.In accordance with the researched robot platform and anti-disturbance performance requirements,based on the combination of state and parameter multi-scale online estimator and linear time-varying model predictive controller,a trajectory tracking adaptive robust control framework is proposed for parameter and load disturbance-oriented in electro-hydraulic actuator.In addition,based on the model prediction method,a time-sharing anti-disturbance control for trajectory tracking is proposed.And further,combining model prediction and sequence optimization whole-body control,a high-speed motion trajectory tracking anti-disturbance method is refined with considering the leg dynamics.Finally,based on the direct collocation method and the segmented expectation curve,a high-precision online optimization method for the trajectory is summarized.The main contents are as follows:1.Under limited sensor configuration conditions of electro-hydraulic actuator,aiming at the three problems of time-varying parameters and load estimation,unmeasurable state estimation and measurable state measurement noise filtering,based on the mechanism model parameters,the parameter sensitivity and fast and slow change characteristics are analyzed.A data-driven state and parameter multi-scale online joint estimation algorithm(MEKF)is proposed with fast-changing and slow-changing time scales.Facing the demand for adaptive robust high-precision trajectory tracking,combining MEKF and linear time-varying model predictive control,the adaptive robust control framework is proposed,and the effectiveness has been verified by experimental research.2.Based on the requirements of high-speed dynamic walking motion and disturbance rejection control of hydraulically driven biped robot,the motion mechanism of underactuated biped robot is analyzed,and the control of biped robot is divided into support phase and swing phase control.In the support period,based on the single rigid body model of floating base,the model prediction controller was used to obtain the optimal contact force spin of the support leg to control the posture and height of the robot through optimization solution.In the swing phase,the forward and lateral velocity of the robot is controlled by calculating the footing point based on the inverted pendulum.A trajectory tracking anti-disturbance time-sharing control framework based on the model prediction method is proposed.Under this control framework,the left and right legs alternately reciprocate,so as to realize the three-dimensional omnidirectional control of the robot.Based on event and time mixing,single-leg dynamics model and joint drive information,the finite state machine principle and foot contact force estimation algorithm are described.Finally,the dynamic anti-disturbance movement of biped robot under high-speed movement,external impact and uneven ground is realized.3.In view of the problem that the leg dynamics are not considered in the anti-disturbance time sharing control framework,based on zero-space projection,weight layering and sequence optimization,the trajectory tracking whole-body control methods are studied respectively.And the dynamic motion and anti-disturbance performance of the methods are discussed and compared through simulation verification.In order to further improve the performance of high dynamics and anti-disturbance,considering leg dynamics and Combining model prediction with sequence optimization whole body control,a high-speed motion trajectory tracking method is proposed,and the high-speed and robust motion against disturbances are verified by simulation.4.Aiming at the problem of trajectory on-line optimization for biped robot from the initial state to the target state under disturbances,based on the single rigid body model and the Hermite-Simpson direct collocation method,the trajectory optimization problem is researched into a finite-dimensional nonlinear programming problem.In addition,the smooth curve segmentation between the initial state and the target state is regarded as the expectation of trajectory optimization,which shortens the time of trajectory optimization and greatly reduces the dimension of optimization variables.Then a high-precision online trajectory optimization method based on direct collocation and segmented expectation curve is proposed.Finally,the effectiveness of the proposed algorithm is verified by simulation.Based on the research accumulation and engineering iteration of the leg robot,as well as the hardware conditions support of the laboratory,the robot topology structure,hydraulic drive system with high power density and airborne control system design were elaborated in detail,and the experimental platform of the first generation hydraulically driven biped robot has been developed.