Gait Optimization Control of Robots Based on Model Predictive Control

With the progress of computer technology, robots can complete a variety of tasks in complex environments, but the existing control methods of multi-legged robots have problems such as insufficient stability and weak gait control ability. Therefore, an optimal control method of robot gait based on model prediction is proposed. In this method, the inverse dynamics model of the robot is constructed, the variational method is used to linearize the model in the model prediction controller, friction constraints are constructed, the ground contact detection is carried out by Kalman filter, and the adaptive adjustment strategy of the robot landing point is established. The experiment shows that the total plantar force of the four legs of the robot is basically the same as the total weight of the robot. When the lateral disturbance occurs, the force distribution between the two legs of the robot is more uniform, and the corresponding load can be shared to maintain its stability. When walking on the seesaw, the distribution of the robot’s planar force can effectively correspond to the state of the fuselage, and the Pitch Angle is between 0.038rad-0.046rad. When measured, the maximum Z -axis force of the planar force is near 80 N, and the minimum planar force is $\mathbf{0}$ when passing obstacles. It can be concluded that the research can effectively control the gait of quadruped robot and improve its task execution ability in rough terrain.