State Estimation Of Quadruped Robot Based On Invariant Extended Kalman Filter

In recent years,with the advancement of my country's modern industry and information technology,quadruped robots have been widely used in industrial development.Compared with common crawler-type and wheel-type robots,foot-type robots can better adapt to different complex scenes and easily overcome obstacles in travel.For the motion control of the foot robot,there are three main methods,including force control,position control and sensor hybrid control.The accuracy and stability of sensor data is a necessary condition to ensure the smooth progress of the above three motion control methods.Based on the constant expansion Kalman filter,this paper creates a state observer for the quadruped robot platform developed in the laboratory,and realizes the precise control of the quadruped robot by fusing the data of the airborne inertial measurement unit and the leg joint sensor.This paper designs experiments to verify the various control indicators of the observer algorithm,including convergence and robustness.The specific research content is as follows:First of all,this article defines the relevant state quantities for the quadruped robot platform developed in the laboratory,introduces the composition of the sensor system,studies and analyzes the method of obtaining the motion state of the robot body by a single sensor,and uses the extended Kalman filter to fuse the data of the dual sensor,and finds the extension The Kalman filter is not suitable for laboratory robot platforms,and puts forward requirements for the fastness of the filtering algorithm.Then,design the control state quantity for the quadruped robot,derive the continuous-time quadruped robot system dynamic equation,and obtain the invariant extended Kalman filter framework.Estimate the deviation of IMU,and derive the augmented invariant extended Kalman filter with the deviation term,analyze the observability of the right invariant extended Kalman filter,and design the filter algorithm to realize the state estimation of the quadruped robot.Finally,through design experiments,the convergence performance and robust performance of the invariant extended Kalman filter are analyzed in detail,and the influence of different covariance matrices on the state estimation of the quadruped robot is explored at the same time.In terms of convergence,the attitude angle and motion speed of the robot state in the filter have good convergence;and for the robustness of the filter,the attitude angle and motion speed of the robot state are also very robust.Good;for the choice of the initial covariance matrix,experiments have found that whether the choice of the initial matrix is a matrix or an identity matrix,its influence on the state estimation can be ignored.