Research On Autonomous Navigation Technology Of The Quadruped Robot For Wild Unstructured Environment

With China’s stride into the era of Industry 4.0,special robots,especially quadruped robots,are rapidly developing in emergency disposal,rescue and relief,field engineering,military reconnaissance support and other special fields under the traction of application demands.The complexity of the application scenario also puts forward higher requirements for the autonomous motion and exploration capability of the robot.The current research on autonomous navigation of quadruped robots has not fully exploited the autonomous motion capability of robots,and there is still room for improvement in the framework of autonomous navigation and path planning algorithms of quadruped robots.Based on this status,this paper,based on the electro-hydraulic hybrid drive quadruped bionic robot developed by the group,conducts research on the autonomous navigation of quadruped robots in unstructured environments,and the specific work is as follows:(1)For the demand of autonomous navigation function of the quadruped robot in wild unstructured environment,a navigation framework based on goal-oriented real-time local map building and planning is built.Firstly,the hardware system of the quadruped robot autonomous navigation system based on solid-state LIDAR+fiber-optic Jetlink combined navigation system+embedded industrial computer is built.Secondly,the software system based on a five-layer model of hardware layer,operating system layer,software platform layer,algorithm layer and application layer is designed.Finally,the overall scheme of the autonomous navigation system is experimented by simulation platform and prototype platform to verify the effectiveness of the framework,which achieves the autonomous exploration function of the quadruped robot without the a priori map of the environment.(2)For the problem of the quadruped robot pose estimation system and environment modeling in complex environments,a quadruped robot pose estimation framework based on multi-source information fusion is constructed,and a study of passability assessment method based on geometric features is conducted based on the pose estimation results and real-time point clouds.Firstly,an odometer based on the solid-state LiDAR-combined navigation system is designed,and the state error Kalman filter is used to solve the equations of motion and observation equations of the system,and the iterative Kalman filter is applied to update the state.Secondly,a local dense elevation map centered on the quadruped robot is constructed based on the point cloud information of the environment acquired by the solid-state LiDAR loaded on the quadruped robot and combined with the posture estimation results.Then,by extracting the geometric features such as slope,roughness and height difference of the local environment,a feasible domain evaluation formula is constructed to obtain an elevation map with feasible domain information,and an occupancy raster map is constructed,and the occupancy probability value of the map is updated by using ray tracing method and based on the real-time environmental information.Finally,relevant experiments are conducted on the simulation platform and the prototype platform to verify the effectiveness of the posture estimation algorithm,the environment modeling algorithm and the Finally,the validity of the pose estimation algorithm,environment modeling algorithm and passability judgment algorithm are verified in the simulation and prototype platform.(3)For the study of the quadruped robot path planning system,a goal-oriented quadruped robot path planning system based on goal guidance is constructed.Firstly,in order to obtain the local optimal target point,this paper introduces the target point as a guide based on the local boundary region exploration strategy of computer vision.Secondly,the path planning algorithm RRT*is improved based on the occupied raster map containing feasible domain information,and a Bessel curve is used to smooth the path considering the motion stability of the quadruped robot to obtain a passable path from the starting point to the local target point.Then,the path points are selected according to the hardware constraints of the quadruped robot itself,and a path tracking controller is designed based on the positional error model of the quadruped robot.Finally,path planning experiments are conducted on the simulation platform and the prototype platform to verify the effectiveness of the autonomous navigation system.