Research On SLAM Of Snake Robot Based On Monocular Vision

The snake robot is a multi-joint,high-redundancy bionic robot.It has a few advantages such as its multi-gait movement capabilities and its adapting to narrow environments,thus it's widely used in disaster relief,industrial detection and underwater survey and other application scenes where human can not reach.In the above environments we cannot accurately obtain the robot's pose and surrounding environment information,Simultaneous Localization and Mapping(SLAM)become a technical bottleneck restricting the autonomous execution and the development to intelligence of the snake robots.This paper focuses on snake robot system modeling and its path tracking methods,it also studies the problems of monocular SLAM such as its scale uncertainty,the inability to rebuild dense 3D scenes,its loss of visual tracking due to dynamic motion characteristics and low light application environment.The main work and innovative content are as follows:(1)Having studied the structure and movement characteristics of the snake,an amphibious snake robot is developed with three-dimensional motion gait which is integrated inertial measurement unit(IMU)and focus stack data acquisition system.Its kinematics model was established following the establishment of the coordinate system of the snake robot SLAM system.The stability control method was explored through mechanical dynamics simulation.(2)As to monocular SLAM scale uncertainty,a high-accuracy passive vision sensor technology based on focused stack light field computation imaging technology is proposed.On the base of the analytic reconstruction and iterative reconstruction,scene light field,depth and full-focus image reconstruction are achieved.The depth of the reconstruction can reach to a centimeter-level accuracy;thus,the problem of scale uncertainty is solved.(3)The traditional monocular SLAM is unable to reconstruct a practical dense 3D scene,while in this paper the depth and full-focus map obtained by the imaging of the focal stack are calculated,and the reconstruction of high-precision,high-resolution,and dense3 D scenes within the field of view(FOV)of a snake robot is realized.The experimental results show that under the same scene,the number of pixels of point cloud using only one tenth image frames in our light field computation imaging which is 7 times of that in the traditional monocular SLAM,thus the dense point cloud of the scene can be reconstructed.(4)In this paper,IMU is used as an auxiliary sensor for pose estimation because of the low-angle,dynamic motion characteristics and low light application environment of the snake robot leads to the loss of visual tracking.By studying the motion constrained features and error propagation characteristics of the snake robot,a snake robot path-tracking method was proposed which aims to control the zero-speed and zero-angle speed constraints in milliseconds.The position error of the path-tracking algorithm was within four percent of the total mileage.What's more,an adaptive dynamic matching IMU and camera combination pose solving method is proposed to solve the problem of loss of visual tracking due to low viewing angle and dynamic motion characteristics,therefore the accuracy of path tracking is further improved,and the position error is reduced to one percent of the total mileage.Based on the development of the snake robot platform,the stability control method of the snake robot was obtained by establishing its kinematics model and mechanical dynamics simulation analysis;the problem of the traditional monocular SLAM scale uncertainty and the inability to reconstruct the dense three-dimensional scenes are solved by reconstructing the light field with the focus stack;the problem of loss of visual tracking due to low viewing angles,dynamic motion characteristics and low light application environment of snake robots is solved through IMU-assisted camera pose solving.The research content of this paper has scientific value in monocular visual scene reconstruction and dynamic environment visual tracking field,and has practical value in snake robot disaster rescue,industrial detection and other applications.