Development Of Robot System And Model Predictive Control For Wheeled Bipedal Robots

Wheel-Legged hybrid robot,which combines the advantages of both wheeled and legged robots by adding wheels at the end of the foot,has developed rapidly in recent years.This paper mainly develops a wheeled bipedal robot,which is studied from the following four aspects.In the overall scheme design,adhering to the principle of minimization of the structure of the robot,according to the movement task requirements of the robot,determine the degree of freedom configuration and related dimensions.In the aspect of kinematics,the forward kinematics is derived,the transformation relation between joint space and Cartesian space,and the transformation matrix between the coordinate systems of each link are established.In order to make the robot balance at first,a simple 2-D wheel-inverted pendulum model was established.In order to realize more complex motions,a full rigid body dynamics model with integrated equilibrium constraints is established.The mechanical system and electrical system of the robot are designed.In order to reduce the weight of the thigh connecting rod,the motor of the knee joint is mounted at the body and the power is transmitted through the synchronous belt.In the electrical system,the industrial computer is used as the controller to send CAN commands to the motor and receive data of IMU through two USB interfaces.A local area network is formed through the airborne router to conduct multi-machine communication through the ROS system,and the robot status display and user instructions are issued on the user’s computer.According to different simplified dynamics models,two control methods are designed,one is pole assignment method based on wheel inverted pendulum model,the other is model predictive control method based on full rigid body dynamics with integrated equilibrium constraints.In the latter,the force balance equations of thighs and shanks were directly integrated into the full rigid body dynamics model,and the obtained dynamics model with balance constraints reduced the model size and retained the freedom of motion of the robot trunk to the maximum extent.Based on the proposed model predictive control method,the control problem of the desired trajectory at the discrete time points of the predicted period is transformed into a quadratic programming optimization problem,and the balance and motion tasks of the robot are accomplished through the coordination of hip joint,knee joint and wheel joint.In order to verify the effectiveness of the proposed control strategy,three scenarios were designed in this paper: shock resistance experiment in situ equilibrium,complex terrain experiment,jumping up stairs and in situ jumping experiment.The final results show that the control strategy des igned in this paper has strong stability and robustness.