| Mobile robot is an important member of the robot family.Positioning and navigation of mobile robot is the key technology to realize intelligent and autonomic robot.Vision-based positioning and navigation can get rich information and has low energy consumption,so it has become an important research direction of mobile robot navigation.In visual positioning and navigation technology,the finite visual field of single camera limits the moving range of the robot,which affects the prediction and pre-adjustment of robot path planning.Enlarging the field of view will decrease the resolution of local details on the image,and affect the accuracy and reliability of positioning calculation.Therefore,it is difficult to combine large field observation with high definition detailed observation.To solve this problem,this paper designs a double camera system with a long focal camera and a short focal camera.The short focal(fish-eye)camera is used to observe the wide range of environment and tracking the target.The long focal camera is fixed on the PAN/TILT Head and used to amplify the target details imaging,obtaining high resolution image of the mobile robot,and then precisely calculate the position and orientation of robot.The dual-camera system is built in practice,and the Billy quadruped robot platform is used to carry out physical experiment verification,and an improved Bi-Directional Rapidly Exploring Random Tree Algorithm is proposed to carry out simulation experiment on the path planning of quadruped robots in static and dynamic environments.The main research contents of this paper are as follows:(1)Designed and developed the host computer software of android for the small electrically driven quadruped robot.Designed and built the external environment perception system of the quadruped robot,combined with visual,auditory and tactile perception.Developed the functions of target following,automatic charging and visual obstacle avoidance,and enriched the human-computer interaction functions of the robot.(2)Using a digital servo and HD camera to build the dual camera system,carry out the kinematics model for the master slave system of dual cameras,calibrate and measure the spatial position relationship of each coordinate system.(3)Using the weighted RGB color histogram based particle filter algorithm to track moving target in the fisheye camera image.Carry out the least square trajectory prediction algorithm based on the historical motion data,the trajectory of the moving target is predicted when it is heavily occluded for a short time.Carry out the uniform sampling method to re-track the lost target.(4)PAN/TILT Head kinematics solution and long-focus camera QR code recognition.According to the 3d reconstruction results of the fish-eye camera,inversely solve the biaxial Angle of the PAN/TILT Head with its kinematics model and rotate the long focal camera to aim at the target area for magnification imaging.Using OpenCV contour detection to identify the QR code on the back of the robot,the center and orientation of the QR code are calculated,and the pose data of the robot are calculated according to the spatial relationship.(5)Improved the RRT algorithm to further shorten the path by the shorten algorithm.Aiming at dynamic obstacle environment,carry out the fast generate algorithm based on probability weights sampling to calculate feasible paths.Through the branch cut calculate,combined with obstacles moving direction and heuristic search,take out the reselect of the path when the obstacles hindered robots current paths,and obtain a new feasible path.Carried out the simulation experiment on the path planning. |
PDF Full Text Request