Underwater Positioning System For Amphibious Spherical Robots

With the rapid development of robotic technologies,there are more and more applications of underwater robot in the field of ocean exploration,search and rescue at sea.Aiming at accomplishing the tasks of tracking and investigation of offshore creatures and offshore self-searching-and-rescuing,we proposed a small amphibious spherical robot system.However,the small amphibious spherical robot system spend more time and expend more energy for accomplishment of wide range of operations,for which mutual tracking and specific formation of multiple robots show dramatic significance and research value.Based on this study,underwater localization of the robots become one basic and essential technology of those in multi-robots cooperation.In this paper,we proposed an amphibious spherical robot system based on visual and inertia sensing modules to meet the requirement of low power consumption,high mobility and high concealment for our self-developed amphibious spherical robot system.The main research contents are as follows:Firstly,to meet requirements of the underwater positioning system,three structural improvements and system design were proposed to improve the underwater adaptability of amphibious spherical robots.The design of the model and parameters were made and optimized by using SolidWorks software.And the feasibility of the design was verified by simulate the robotic movement.Finally,a new amphibious spherical robot prototype was made.Secondly,taking into account the requirements of low power consumption,high mobility,high hidden for amphibious spherical robots,the existing underwater robot technology have many problem,such as complex system,high cost,and poor positioning accuracy.The underwater visual positioning system of amphibious spherical robot based on the color image and the depth image of the ToF camera was proposed.Firstly,the image plane localization method based on HOG and ColorName features was studied.The traditional HOG and ColorName feature were fused to describe the underwater amphibious spherical robot,and the recognition effect was greatly improved.Then,the 3D position reconstruction method based on ToF camera imaging model was studied,and the target depth information was obtained from the depth image of the ToF camera,which solved the problem of distance acquisition of visual location.Finally,the error of the camera in water was analyzed,and the corresponding error correction scheme was designed for the image distortion and depth error respectively,to improve positioning accuracy.Thirdly,aiming at the problem that the underwater visual positioning range which is limited by the camera angle and the effective depth,a multi-angle underwater visual positioning method based on the MEMS pose information of the MEMS inertial sensor was proposed.On this basis,a multi-robot underwater positioning system was proposed by normalizing the position of each robot.Firstly,the robotic coordinate system based on the Euler angle transformation matrix was studied.By using the advantage of the underwater rotation of the sphere structure,the problem that the visual orientation is limited by the camera angle was solved.Then,the relationship between the positioning coordinates of multiple robots was studied.The multi-robot absolute coordinate normalization method was used to fuse the multi-robot positioning information to solve the problem that the visual localization is limited by the effective depth.Finally,the error of the sensor was analyzed,and the Kalman filter was designed to eliminate the error to improve the system stability.