| In order to alleviate traffic congestion,reduce the incidence of traffic accidents and improve road utilization,the development and application of autonomous driving technology has become a key direction of research in today’s automotive industry.Despite the achievements of autonomous driving technology,there are still many problems in dealing with the complex urban road environment,and the problem of environment perception and driving strategy in the overtaking process is one of them.Therefore,the study of overtaking strategy for self-driving vehicles in urban road environment has certain theoretical significance and application prospects.In this paper,we used camera as vehicle sensor and solved the problem of insufficient perception range of fixed camera view based on active binocular vision.Combining the position and speed information of surrounding vehicles,we divided the overtaking scenario and established the decision model of overtaking behavior.We planed the overtaking lane change trajectory and tracked the trajectory,and designed the real vehicle test to verify the research content of this paper.The main research contents of the paper include:(1)An active binocular vision obstacle detection algorithm based on VIDAR(Vision-IMU based Detection and Range Method)was proposed.In this paper,the VIDAR obstacle detection algorithm was used to make preliminary determination of generalized obstacles,and the obstacle search mechanism based on binocular non-matching area was established according to the area where the obstacle was located.In order to achieve binocular ranging and speed measurement of obstacles,this paper designed the rotation algorithm and used the normalized mutual correlation template matching method to track the obstacles.YOLO v5 s combined with traditional binocular vision was used to compare the detection effect with this paper’s method,and the results proved the superiority of this paper’s method in terms of detection range and detection accuracy.(2)Active binocular vision-based decision making for overtaking behavior was investigated.The overtaking scene was simplified and modeled,monocular and binocular vision conversion and camera rotation strategies were designed,and an overtaking environment perception model based on active binocular vision was constructed.For the straight-line overtaking stage,the target was tracked using Kalman filter and the distance measurement was performed using the small-aperture imaging principle.The minimum safe distance of lane change for the left lane change phase and the right lane change phase of the vehicle was calculated for the three cases of possible collision.Using a finite state machine,a decision model of overtaking behavior was established to determine the driving strategy of the self-car in different overtaking scenarios.(3)Planning and tracking control were performed for the overtaking lane change trajectory.A transverse-longitudinal trajectory planning model based on the Frenet coordinate system was developed using a quintuple polynomial method.The set of trajectories in the transverse and longitudinal directions was obtained according to different vehicle target states and different sampling times.And the evaluation function and acceleration and curvature constraints were designed to filter the trajectories.The model predictive control algorithm was used to track the lane change trajectory,and the joint simulation based on Car Sim and MATLAB/Simulink was performed.The results verify that the controller has good tracking effect.(4)Experimental validation was conducted for the overtaking environment perception model and the overtaking behavior decision model.The active camera head was designed and the test platform was built,and the control of the camera turning angle was realized by using the motor and STM32F103VET6 development board,and the parameters of the test vehicle were collected by using the GPS-IMU module.Using the modified test vehicle,four different scenarios were tested.The test results demonstrated the feasibility and effectiveness of the active binocular vision-based lane change overtaking strategy studied in this paper. |
