Research On Autonomous Gait Planning Technology Of Quadruped Robot Based On Local Map

In the field of mobile robot research,due to its superior terrain adaptability,legged robots have more application prospects in military operations,disaster relief and other fields than wheeled and tracked robots.For this reason,they have been widely concerned by experts and scholars at home and abroad.Regarding the walking of quadruped robots,the current research focus is mainly on the use of advanced sensor technology to further improve the adaptability of robots to complex and unknown environments.To this end,starting from the existing quadruped robot motion control method and experimental platform of the research group,the existing static gait planning method is improved to enable the robot to pass through more complex and extreme terrain;RGBD camera is introduced to complete the autonomous positioning and local map construction work of the robot,which enables the robot to autonomously plan the step height and step length of the static gait with the help of terrain information to pass complex and unknown terrain.The main contents are as follows:(1)A quadruped robot model is constructed,and the coordinate system of the single leg model is established.The D-H parameter method is used to model and analyze the robot single leg,and the forward and inverse solution equations of the robot single leg are given.In view of the shortcomings of the existing intermittent static gait when facing high obstacles or large slope terrain,the stability of the robot when facing complex terrain is enhanced by changing the step sequence and the center of gravity adjustment strategy.Finally,the effectiveness of this method was verified by the experiment of climbing a 60-degree slope based on Webots simulation environment.(2)Visual and inertial sensor modeling analysis and calibration.The complete coordinate system of the robot pose estimation system is constructed,and the camera model and IMU measurement model are given.The noise of the IMU is analyzed based on the Allan variance analysis method.The internal parameters of the camera and the extrinsic parameters of Camera-IMU are calibrated using the kalibr toolbox.The noise and offset of the IMU are calibrated by Allan analysis of variance.(3)Aiming at the problem of visual inertial odometer positioning,the first feature tracking scheme based on optical flow method is used for visual front-end processing.Among them,the ORB operator that takes less time is selected as the key point for feature tracking;in the IMU data processing stage,pre-integration technology is used for the preliminary processing of IMU data,derivation of IMU continuous time error transfer equation and IMU pre-integration update method.For the nonlinear problem of visual inertial information,the visual measurement information,IMU measurement information and the information after marginalized old frames are used as constraints to construct the least squares problem,which can be solved iteratively using the LM method,and finally the pose estimation of the robot is obtained.(4)Use the calibrated camera internal parameters to complete the conversion from depth image to 3D point cloud,and reduce the number of point clouds by using voxel filters and save the time of point cloud processing.Derive the variance of grid height value and grid position uncertainty covariance,which improves the accuracy of the map.For the calculation of each grid height value,all grid heights within its confidence ellipse are weighted and fused,and finally a local map of the robot is constructed.(5)Use ROS and Gazebo to build a simulation environment,which uses ROS to design a complete set of systems,including robot motion control system,sensor system,robot positioning system and local map construction system.In the simulation environment,the visual inertial odometer positioning experiment,the local map construction experiment and the autonomous gait planning experiment are completed respectively.The simulation shows that the autonomous gait planning system of quadruped robot based on local map designed in this paper can improve the adaptability of quadruped robots to complex environments.