Research And Implementation Of Motion Stability Of Legged Robots Based On IM

Compared with wheeled and tracked robots,legged robots can move smoothly in whether flat or rough terrain relying on the discontinuity of the contact point on the ground.But the discontinuity of the contact point also leads to the poor motion stability of the legged robot.How to maintain the motion stability of a legged robot in a complex environment is a problem blocking the way of legged robot being widely used.To solve the above problem,this paper studies the walking stability control method of the quadruped robot,designs a piecewise cycloid foot trajectory equation with variable shape parameters,and compares the piecewise cycloid trajectory and the cycloid trajectory in Webots dynamic simulation environment.The influence of piecewise cycloid trajectories on quadruped robot motion stability in terrains with different slopes is studied,analyzed and tested.To further confirm the experimental conclusions obtained in the dynamic simulation environment,this paper designs and builds a quadruped robot,conducts experiments and analyzes the experimental results.In order to further enhance the motion stability of the piecewise cycloid trajectory for the quadruped robot under terrain with different slopes,reinforcement learning is used to optimize the shape parameters of the piecewise cycloid trajectory,such as step length and step height.Piecewise cycloid with fixed shape parameters and optimized shape parameters are used in comparative test to analyze the influence on quadruped robot motion stability in terrains with different slopes.The main research contents of this paper are as follows:(1)The forward and inverse kinematics analysis and foot force analysis of the parallel legs of the quadruped robot are carried out with geometry relationship.The 3D model of parallel leg quadruped robot is optimized for building robot dynamics simulation model in Webots environment.And a quadruped robot is built for further experiment.(2)A piecewise cycloid trajectory is designed to improve the quadruped robot motion stability in horizontal ground,uphill and downhill terrains based on cycloid trajectory.-10°~10° slope terrains are built in Webots environment to conduct the comparative experiments between piecewise cycloid trajectory and cycloid trajectory.Simulate IMU sensor in Webots environment are used to acquire the pitch and roll angle for motion stability evaluating indicator.Experiment results show that the piecewise cycloid trajectory reduces the pitch and roll angle amplitude by 17% and 27% during the quadruped robot’s walking process,and the pitch and roll angle variance is reduced by 42.2% and43.3%,compared with the cycloid trajectory with same shape parameters.The quadruped robot is used for further verification on the slope in range of-8.7°~7.6°.The BNO080 IMU sensor is used to obtain the pitch and roll angles of the quadruped robot body,and the motor current is used to estimate the foot torque.The experimental results show that compared with the cycloid trajectory,the pitch and roll angle variances of the piecewise cycloid trajectory on the horizontal ground and downhill terrain are reduced by 13.8% and 43.5%,and the variance of the foot moment is reduced by 34.6%.The variance of foot-end moment in uphill terrain is reduced by 65.8%.Under different slope angles,the influence of piecewise cycloid trajectory on walking stability varies greatly.(3)To further improve the motion stability of the quadruped robot with piecewise cycloid trajectory,reinforcement learning is used to obtain the optimal shape parameters of the piecewise cycloid trajectory under different terrain slope angles.According to the shape parameters of the piecewise cycloid trajectory,the state space,action space and reword function in the reinforcement learning controller are designed to build the reinforcement learning controller.(4)The optimal shape parameters of the piecewise cycloid trajectory corresponding to different slope angles are learned through the DDPG algorithm in the Webots simulation environment using the reinforcement learning controller.Comparing the piecewise cycloid trajectories with optimal shape parameters and fixed shape parameters under different slope angles,the experiment results show that,the piecewise cycloid trajectories using the optimal shape parameters are better than those with fixed shape parameters,compared with the piecewise cycloid trajectories with fixed shape parameters.The pitch and roll angle variances of the quadruped robot walking on the horizontal ground are reduced by98.1% and 81.2%,and the pitch and roll angle variances on the downhill ground are reduced by 97.3%and 58.5%.The pitch and roll angle variances are reduced by 16.6% and 22%.