Research On Motion Control Algorithm Of Quadruped Robot Based On Trajectory Optimization

Quadruped robots have significant advantages over other types of robots,both in terms of unstructured terrain adaptation capability and highly dynamic motion.Relying on these two advantages,quadrupedal robots have a wide range of applications and research value in the current social field.Since the quadruped robot system is underdriven,it is difficult and challenging to achieve its high dynamic motion capability.Therefore,this thesis presents an in-depth study on the modeling methods of kinematics and dynamics of quadruped robots,algorithms for estimating motion states and terrain conditions,dynamic real-time motion control algorithms,and gait adjustment strategies.The details are as follows:(1)By constructing the motion tree of the quadruped robot and establishing the coordinate system of each joint of the quadruped robot,the algorithm of the kinematics of the quadruped robot and the set of kinetic equations of the Newton Euler iteration of the quadruped robot based on space vectors are completed.(2)The kinematic model is used to establish the set of Kalman filter equations to complete the estimation of the states of the center of mass position,velocity and footend position of the quadruped robot; the planned gait phase and the knee joint moment of current feedback are fused into the Kalman filter to complete the estimation of the real contact state.The gait switching strategy based on the event trigger mechanism is designed to complete the switching of the actual motion state of the leg of the quadruped robot.For the judgment of early touchdown,the switching from moment feedback to contact state lasts about 15 ms,which is about 34 ms earlier than the switching from swing phase to support phase in normal gait planning.(3)Establish the state space with the dynamics model variables as the states,complete the linearization of the rotation matrix and angular velocity based on the variational method,and do the vectorization,and obtain the optimal state and control quantities after the quadratic planning solution; the optimal control quantities are transformed into the control of the joints through the joint space of the legs.Finally,the position error on x-axis and y-axis is 3cm,and the z-axis position error is 7cm; the x-axis velocity error is 0.1m/s,the y-axis velocity error is 0.04m/s,and the z-axis velocity error is 0.2m/s.(4)Based on the characteristics of the ZMP,the position of the ZMP of the quadruped robot is planned to be kept on the diagonal of the support leg.Considering different situations,the planning of the offset of the landing point of the foot end under the diagonal gait motion of the quadruped robot is completed to ensure the stability of its motion.A judgment equation to test the stability of the diagonal gait motion of the quadruped robot is established,and the distance from ZMP to the diagonal leg support line is reduced by 2cm on average after adopting this strategy than before.(5)According to the algorithm and strategy designed above,the framework of the quadruped robot system as a whole is constructed,and the scheduling process of the overall algorithm is established; the estimation of the contact state,the stability of dynamic motion control and the gait adjustment strategy are tested under the scenes of stairs,grass and gravel road by the simulation platform and the real physical platform to verify the applicability and effectiveness of the algorithm and strategy proposed in this paper.