Research On Target Detection And Location Grasping Of Indoor Mobile Robots

At present,the proportion of the elderly population and the number of disabled people in China are increasing year by year,and the burden of social elderly care and disability assistance is becoming increasingly heavy.How to improve the quality of life of elderly and disabled people with mobility difficulties and reduce the burden on young people is one of the urgent social problems to be solved.With the development of robotics technology,using home service mobile robots to solve the above problems has become a feasible solution.This article mainly studies the autonomous navigation and positioning grasping control technology of indoor service-oriented mobile robots,and constructs a mobile robot movement grasping control system based on the ROS framework.The specific content is as follows:(1)According to the functional needs of home service mobile robots,corresponding hardware selection was carried out,and the overall software system of the mobile robot was constructed based on the ROS framework;The kinematics of the robot chassis with four Mecanum wheel is analyzed,and the corresponding relationship between the linear speed,angular speed of the mobile robot and the rotational speed of each Mecanum wheel is established;Compare and analyze the current mainstream two-dimensional laser SLAM algorithms and determine through comparative experiments that the Cartographer algorithm is the mapping algorithm of this article;An indoor navigation obstacle avoidance system framework was established by analyzing and comparing the current mainstream path planning algorithms,with A * algorithm as the global path planning and dynamic window algorithm as the local path planning.The effectiveness of the framework was demonstrated through specific navigation experiments.(2)After careful analysis and comparison of the current mainstream object detection algorithm framework,YOLOv5-S algorithm was determined as the basic framework of the object detection algorithm in this study;The defects in YOLOv5-S algorithm are emphatically analyzed,and the Leaky in the original algorithm is replaced by the SMU activation function＿Relu activation function replaces GIOU loss function in the original algorithm with CIOU bounding box regression loss function;To further improve the robustness of the algorithm in indoor object detection of mobile robots,an indoor scene dataset containing a large number of motion blurred images and semi occluded images of objects was created to complete the training task of the algorithm model;The experimental results show that the target detection algorithm model of this study is applied to the dataset produced in this study m AP@0.5 The accuracy index result of 0.95 is 85.4,which is an improvement of 8.1 compared to the original YOLOv5-S algorithm.The speed FPS index is 65.36,which can meet the requirements of real-time performance.(3)Based on the principle of Kinect V2 camera,a conversion formula was derived from the two-dimensional pixel coordinate value of the center point of the target detection frame to the three-dimensional coordinate value of the target grasping point robotic arm in the world coordinate system.The internal and external parameter matrices in the conversion formula were solved using Zhang Zhengyou’s chessboard calibration method and hand eye calibration method;In order to use the conversion formula and improve the accuracy of the conversion results as much as possible,three preprocessing tasks were carried out on the images collected by the Kinect V2 camera.Firstly,RGB images were registered to correspond one by one with the depth image pixels.Secondly,three improved filtering methods were proposed to address the problem of missing depth values at the capture points.Based on the experimental results,an improved mean filtering method was chosen to solve the problem,Finally,the problem of RGB image distortion was analyzed and specific correction formulas were elaborated;The experimental results show that the MSE result of this conversion formula in the range of 0.5m to 3m is 174.7,and the average absolute error result is10.2,which has more accurate positioning accuracy.(4)The forward and inverse solutions of the kinematics of the manipulator are studied,and the joint coordinate system of the manipulator and the corresponding DH parameter table are established;A detailed analysis of the geometric structure of the robotic arm was conducted,and the difficulty of reverse solving was reduced by decoupling the robotic arm into three parts.The calculation formulas for the motion of each joint of the robotic arm were derived,and the correctness of the formulas was verified through corresponding testing experiments.Finally,a robotic arm grasping control experiment was conducted,and an average grasping success rate of88.3% was obtained,Proved the effectiveness of the grab control scheme proposed in this study.