Research On Multi-Object Tracking And Vehicle Steer Control Of Unmanned Driving System

With the improvement of people’s living standards and the rapid development of the automobile industry,the number of cars increases and the following traffic accidents also occur frequently.The purpose of the unmanned system is to replace the function of driver,to release the driver from the dangerous and boring car-driving,and to reduce the possibility of traffic accidents.As an important part of the unmanned system,multi-object tracking and vehicle steer-by-wire can make the unmanned driving car detect and track the object around the vehicle,obtain the object trajectory and predict the next movement.The unmanned driving car can stop or turn according to the trajectory,which ensures the driving safety.Therefore,research on multi-object tracking and steer-by-wire of unmanned driving system has an important theoretical significance and valuable application.Practice shows that deep sort algorithm and sliding mode control are effective means to solve the problems of multi-object tracking and steer-by-wire respectively.However,how to improve the performance of multi-object tracking,solve the problem of object ID(identity)changes in the tracking process,ensure the accuracy of vehicle steering control and the reliability of steering power transmission have become the problems to be further studied and solved.In this paper,the multi-object tracking and steer-by-wire of unmanned driving system are researched:(1)Multi-object tracking requires fast and accurate input.In this paper,yolov4 network is used as object detector,which has the characteristics of high detection accuracy and fast detection speed.Using data set to train the target detector based on yolov4 network,the training and test results show that the object detector can meet the expected performance requirements.(2)Aiming at the problem of object ID change in the original deep sort algorithm,this paper uses Kalman smoothing algorithm to modify the trajectory estimation of the object;The relative motion network model is added to deal with the occluded objects;SIOA(Sum Of Intersection over Target Area)matching algorithm is used to replace IOU(Intersection Over Union)matching algorithm to improve the matching performance;The tracking performance is tested on MOT16 data set,and the results show that the improved deep sort algorithm solve the problem of object ID change,and improves the tracking performance.Finally,multi-object tracking experiments are carried out on campus,and the results show that the improved deep sort algorithm has good tracking performance and practicability.(3)According to the track of the objects and its motion prediction,using the artificial potential field method to plan the reference track.Designing control method to make the vehicle steer according to the expected angle.Aiming at the problem that the control method based on the dynamics of steer-by-wire system is difficult to obtain satisfactory control performance in practical application,an adaptive fuzzy high-order sliding mode control strategy is designed in this paper.The fuzzy logic system is used to approximate the function of steer-by-wire system,and the chattering phenomenon is effectively attenuated by using high-order sliding mode technology.Through the Lyapunov stability analysis and the comparison between simulation and experiment,the rationality and superiority of the adaptive fuzzy high-order sliding mode control method designed in this paper in the application of steer-by-wire of unmanned driving vehicle are proved.(4)When the actuator fails,the steering motor can not generate enough torque to drive the front wheels at its reference steering angle.To solve this problem,a fault-tolerant controller based on adaptive sliding mode is designed in this paper.The adaptive fuzzy system is used to approximate the friction torque and slewing torque of the steer-by-wire system.Through the robust term of sliding mode control,the negative effects of actuator failure,modeling error and external disturbance on the control performance are reduced,The discontinuous switching function of sliding mode control only appears in the time derivative of control signal,which eliminates chattering phenomenon and eliminates the need to increase the higher-order time derivative of sliding mode variable in control design.Theoretical analysis shows that the designed method can achieve the stability of closed-loop steer-by-wire system.Finally,the effectiveness and practicability of the method are proved by simulation and experiment.