Research On Vehicle Active Steering Control Of ADS Based On Vision

In recent years,with the rapid development of the automobile industry and the growth of the world population,people’s demand for vehicles is increasing day by day,and the traffic safety problem is also increasingly serious.Traffic accidents have become the third largest death cause of the world’s population.Among which the traffic accidents caused by alcohol driving,fatigue driving,speeding,lane departure and other reasons account for a large proportion.As an effective technology to improve the driving safety of vehicles,assistant driving system has emerged as the times require.With the increasing market demand,assistant driving system is gradually becoming the necessary application of all kinds of vehicles.Among them,the assistant driving system,which is based on lane keeping and active steering control,has become an important guarantee for the safety and comfort of vehicles in the process of driving.The lane keeping active steering control system designed in this paper can ensure that the vehicle can keep driving in its own lane,and can actively turn to avoid obstacles when there are obstacles in front of it.In this paper,a reasonable and effective control strategy is adopted to control the vehicle,so that the vehicle steering control system studied in this paper can meet the requirements of vehicle ride comfort,stability and safety requirement.In this paper,the assistant driving system based on vision,vehicle active steering control system is designed.It mainly discusses the following technical problems:(1)This paper analyzes the research results of current domestic and foreign scholars in road image processing,vehicle local path planning and vehicle active steering control,and determines the main methods and key technologies used in this study.(2)Using the method of deep learning to process the road image,constructing a model of convolutional neural network for multi-task learning,innovatively designing the network model,optimizing the network structure,adjusting the network parameters.Selecting the data set of the training network according to the research direction,carrying out the network training and parameter fine-tuning after preprocessing the data set image,so as to achieve the optimal effect of the network,so as to complete the task of lane line and vehicle detection.(3)Carry out vehicle local path planning,and the optimal rapidly-exploring random trees are used as the local path planning algorithm in this paper.The algorithm is improved,multi-constraint conditions are added,and the planned path is smoothed by B-spline curve.(4)The lane keeping lateral control strategy is established,the whole vehicle dynamics model is established based on Matlab/CarSim,the lane departure judgment conditions are designed,the lateral lane controller based on model predictive control is established,and the controller designed in this paper is simulated,analyzed and compared.(5)The active steering control strategy based on SBW model is established,the state space equation of vehicle steer by wire system is calculated.The PD control strategy is used to adjust the steering deviation and deviation rate of vehicle steering,the adaptive fuzzy logic is used to compensate the uncertain variables of the system,the stability of the system is analyzed.The SBW control experimental platform is built,which is based on dSPACE real-time system is verified by experiment,and the feasibility of SBW control system based on PD+adaptive fuzzy feedforward compensation control is proved by comparison.