Research And Application Of Indoor Autonomous Navigation Based On Depth Camera

In recent years,with the continuous development of robotics technology,robots have slowly entered people's daily life,helping people to complete some tasks in life.The more intelligent they are,the more important they can play when they are be used.The Robots which can move autonomously could help people accomplish more and more tasks and problems.This paper mainly studies the autonomous navigation of robots indoors.It is mainly divided into Simultaneous Localization and Mapping(SLAM)and path planning,which are the hot-topics of current research.The SLAM describes the surrounding environment for the robot and determines the position of the robot in the environment and the path planning helps the robot to plan the optimal path from the current position to the target position.This paper mainly studies the hardware and software,SLAM system and path planning of the robot mobile control:1)Firstly,the control circuit of the robot mobile chassis is built.The two-wheel differential drive control mode is used to control the robot movement by using the positive kinematics model.The host computer sends the moving speed,which is converted into the left and right wheel speeds of the robot chassis.Finally the PID parameters is debugged so that the robot can move at a steady speed to complete the stability control of the robot chassis.2)The front-end of the visual SLAM is studied.The RGB-D camera is introduced,and the camera model,the internal and external parameters of the camera are analyzed.The internal parameters of the camera are calibrated under the ROS system.The feature points are extracted according to the ORB feature,and feature matching is performed on adjacent frames using Fast Nearest Neighbor(FLANN).Finally,the matched error terms are removed using the RANSAC algorithm.Since the noise of the depth information of the RGB-D camera has a large influence,the spatial projection of the 3D-2D is used to obtain the motion estimation of the continuous frame.The PnP algorithm is studied to solve the six-degree-of-freedom pose of the robot,and nonlinear optimization is used to minimize the error term.3)The SLAM back-end was studied.The motion and observation model of SLAM are introduced,the graph optimization based on robot pose and landmark points is studied,and the pose and landmark points are optimized.The sparse matrix of the back-end optimization solution is analyzed,which provides the basis for the real-time performance of SLAM.This paper studies the loop detection algorithm,builds a dictionary based on the characteristics of the ORB,and uses the word bag model to complete the loop detection to construct a globally consistent environment map and robot movement trajectory.In order to realize obstacle avoidance and navigation functions,it is necessary to construct a dense map and convert it into an octree map,compress the map storage size,and reduce the map storage pressure based on the navigation function.4)The global path planning algorithm is researched so that the robot can complete autonomous navigation under the environment map that has been built.The A^* algorithm with breadth-first search and heuristic cost is mainly studied,and the algorithm is improved to make the robot more stable in autonomous navigation.