Research On Navigation Control Of Self-Reconfigurable Spherical Robot Docking

Resistant capsized and omni-directional spherical robot stable and fast moving ability,the related research in recent years has been a hot and difficult topic in the study of mobile robot,this project for a new type of spherical robot "BYQ-CG" face detection,prevention,checking is given priority to the task of geological disaster can adopt the strategy of "pieces",performed by a single spherical robot task;In the face of the task robot which focuses on rescue and enhancement of passability,the whole robot takes the strategy of "integrating zero into whole" to reconstruct.Therefore,how to help the spherical robot to realize reconstruction and docking is the focus of this paper.For the spherical robot refactoring technology research,the most fundamental and important work is the spherical robot,through the mobile robot navigation and positioning algorithm and realize the pose estimation,such as dynamic target detection precision for the spherical robot refactoring is feasible for docking,but is still a huge challenge,it is also the meaning of what we need to study in this paper.In this paper to reconstruct spherical robot as the research object,the spherical robot to accurately realize the reconstruction of external docking task as the research target,combined with the advantage of multi-sensor fusion to the reconstruction of spherical robot navigation and positioning algorithm and dynamic target detection and position estimation method to study,to ensure the reconstruction in the spherical robot closed spherical shells and spherical shells was able to get open the working condition of the movement of the robust global positioning,and able to be in "coarse alignment" and "accurate docking" stage of subtasks can obtain stable motion target detection and relative pose estimation.The research achievements of this paper are as follows:(1)Design of Positioning,Navigation and Detection System for Self-Reconfigurable Spherical Robot.Based on the analysis of the outdoor application scenarios and the characteristics of the sensors carried by the self-reconfiguring robot,as well as the existing methods of positioning,navigation,target recognition and detection,a reconfiguring docking oriented navigation and detection system for the self-reconfiguring spherical robot BYQ-CG was designed.According to whether the spherical shell is open or not,the spherical robot can be divided into two positioning conditions.The loose coupling inertial system design scheme of global navigation satellite system and the global positioning scheme of differential GNSS integrated IMU and binocular vision are designed respectively.The reconstruction docking task is divided into two target detection sub-tasks,and the target detection and recognition schemes are designed respectively.And through the experimental study of the quality and characteristics of the original information of the sensor carried,the feasibility of the research and the necessity of adopting the research scheme and system design in this paper are explored.(2)Research on Navigation and Positioning Algorithm for Self-Reconfigurable Spherical Robot Docking.In order to realize the global positioning of the self-reconfigurable spherical robot with high accuracy and robustness in outdoor scenarios,and to provide navigation support for the reconfigurable docking,an enhanced adaptive integrated navigation method based on the loose coupled inertial system(GNSS/INS)of the global satellite navigation system was proposed in this paper for the closed condition of the spherical shell.For the complicated working condition of the spherical shell open a visual inertia odometer fusion of GNSS global positioning algorithm,the visual inertia odometer(VIO)pose information fusion and GNSS positioning information,so as to realize the precise local and global without drift real-time joint pose estimation,for the reconstruction of spherical robot with real-time and robust global position information,and add the front characteristic management optimization design,and add the global pose information tightly coupled backend,based on nonlinear optimization with residual vision and IMU residual optimization,add the global pose residuals,The calculation speed and robustness of the fusion positioning system are improved,and the target positioning error in the actual scene is less than 0.5m.Finally,the path planning algorithm based on the fusion of A*and DWA is used to provide the navigation path for the self-reconstructing spherical robot docking(3)Oriented "coarse alignment" subtasks phase in this paper,we design a lightweight fast target detection algorithm,by using ResBlock 3-d feature extraction module to replace the part CSPBlock feature extraction module,and add residual auxiliary module,while further lightweight target detection algorithm to ensure the detection target recognition accuracy,at the same time,through the depth of the fusion methods of the camera's depth measurement information,makes the target detection subsystem has the ability to measure and target distance at the same time.Facing the sub?task of "accurate docking",this paper takes specific Aruco QR code as the target,indirectly calculates external parameters,assists the robot to obtain accurate docking interface pose,and realizes relative pose estimation.(4)For the reconstruction of the spherical robot application scenarios,for the reconstruction of the spherical robot real docking,navigation and positioning algorithm and target detection identification algorithm for the experimental research,and the algorithm and system integration,in BYQ-CG docking test on the robot platform,in order to complete validation in this paper,we study the feasibility and effectiveness of it.