Research On Grasping Control Of Intelligent Vehicle Based On Visual Information

As a robot,the intelligent vehicle has good mobility and adaptability to complex environments,which can replace people to complete many tasks.In real life,intelligent vehicles are usually equipped with some expensive sensors to realize the perception of the surrounding environment,such as GPS navigation and positioning system,lidar,depth camera and so on.However,the increase of the number of sensors will increase the control difficulty of the intelligent vehicle,and the high cost also limits the use range of the intelligent vehicle.In order to solve these problems,this thesis takes the low-cost intelligent vehicle equipped with a manipulator and a camera as the research object,and studies the grasping control of the intelligent vehicle.The specific research contents are as follows: 1.An intelligent vehicle with a four degrees of freedom manipulator and a camera is built,and the communication system between the intelligent vehicle and the computer is built to transmit image information and control commands.The software framework of intelligent vehicle and computer is designed.2.Train a cascade classifier to recognize the target object.Aiming at the problem of poor real-time performance of cascaded classifiers,a cascaded classifier based on a prediction model is proposed.This method reduces the target search area by predicting the motion range of the target object in the image,thereby accelerating the speed of target recognition.Subsequent experiments verify the rapidity of the proposed method in identifying target objects.3.In the depth measurement of the target object,the traditional measurement method based on pinhole imaging principle will produce inaccurate measurement when the camera is difficult to calibrate and correct.In this thesis,a depth measurement method based on neural network is proposed.This method can ensure the accurate measurement of the target depth and avoid the camera calibration.The proposed depth measurement method is compared with the traditional depth measurement method,and the experimental results verify the superiority of the proposed method.4.The single intelligent vehicle grasping control and the two intelligent vehicles cooperative grasping control are studied.Firstly,two neural networks are constructed to calculate the joint angle of the manipulator and the depth of the end of the manipulator respectively.Secondly,PID controller and control scheme are designed to control the intelligent vehicle to search for the target object,and control the manipulator to grasp the target object effectively.Then,the communication structure and corresponding software framework of the cooperative grasping of the two intelligent vehicles are designed,and the control scheme of the cooperative grasping of the two intelligent vehicles is designed.Finally,experiments are carried out to verify the effectiveness of single intelligent vehicle grasping the target object and two intelligent vehicles grasping the target object cooperatively.