Research On Leg Motion Buffer And Balance Control Of Quadruped Robots

In recent years,mobile robots have been widely used,especially wheeled robots.However,the special terrain greatly limits the range of movement of the wheeled robots.In contrast,quadruped robots are more flexible and can pass through a variety of terrain,such as steps,ruins,obstacles and so on.The quadruped robots are capable of disaster relief,mountain transportation,exploration and other application scenarios.Due to more degrees of freedom,the motion control of quadruped robots is more complicated than that of wheeled robots.In recent years,the motion control of quadruped robots has become one of the most popular research topics.At present,most quadruped robots are driven by motors,and the position servo of motors can ensure the robots' high position accuracy.However,when walking and jumping,the quadruped robots will suffer the impact force from the ground because the speed of the foot end landing on the ground is not zero.Due to the high stiffness characteristics of position servo control,the impact force transmitted to the torso will destroy the balance of the robots,and will cause larger leg joint torque response.This kind of torque response leads to abnormal current pulse in the motor,which will consume more energy and even makes the instantaneous current of the electrical system to be too large to trigger the power-off protection.Therefore,the position controlled quadruped robots are difficult to satisfy the requirements of stable motion control.To solve this problem,a motion buffering method of hybrid position and torque control is proposed.When the leg and the ground impact each other,the joint motor uses the torque control mode,and uses the compliance control method to reduce the leg stiffness,so that the leg can absorb the impact force and reduce the joint torque response when landing on the ground.Considering the inertia consumption of the leg,the dynamic model of the single leg is derived.Based on this model,the model predictive control algorithm is applied to achieve better position accuracy under torque control.Trotting gait is the most common gait of quadruped robots.When trotting,the legs of the quadruped robot are supported alternately.How to keep the robots balance is difficult.To solve this problem,a balance controller of quadruped robot is designed.This controller calculates the desire force of the torso by PID controller for its posture and position,and distributes the desire force of the torso to each leg through the optimization algorithm based on quadratic programming,so that the robot has the ability of posture adjustment.Under the action of this balance controller,the robot can adjust the posture of the torso to keep balance through the posture error.A motor drive system based on EtherCAT bus is built on the quadruped robot prototype and the fast switching between position and torque control of the motor is realized.The jumping motion buffering experiment is completed on the a2-DOF single leg platform,which verifies the motion buffering ability and good position tracking performance of the robot.The model predictive control of single leg is realized in the simulation environment,and the position following performance of model predictive control is verified.The experiment of trotting gait is completed on the quadruped robot to verify the balance ability of the robot.