Motion Control Of Collision Avoidance For Intelligent Vehicles In The Specific Road Environments

Technology development for intelligent driving facilitates the upgrade of automotive industries,improves the quality of people's travelling,and handles the problems of traffic,energy,and environment,etc.At present,the key technologies for intelligent driving mainly include sensing,localization,identification,decision making,planning,and control.Because driving safety is vital to road traffic,collision avoidance control is of importance in control technologies for intelligent vehicles.Except for some vehicles for special purposes,most of vehicles travel in different road environments,such as village roads,urban roads,and highways.Moreover,the intelligent vehicle may have different collision avoiding tasks in these different road environments,while one single control strategy can hardly deal with various driving conditions.Therefore,this dissertation focuses on the motion control of collision avoidance for intelligent vehicles on the one-way road in the residential area,the bend road in the suburbs,and the highway road with multiple agents.In these specific road environments,three control systems are proposed to realize the driving safety for intelligent vehicles,which are supervisory assistance control system,safe steering control system,and intelligent hybrid control system,respectively.Yaw motion stability is analyzed based on various conditions.Firstly,model the vehicle motion and obtain the phase portraits of yaw motion using the nonlinear vehicle dynamics model.Secondly,analyze the characteristics of vehicle yaw motion by comparing the phase portraits of different conditions.Thirdly,based on some assumption,the boundary of the stable condition can be solved with the Jacobian matrix.This study provides the vehicle model basis for motion control.The motion control of collision avoidance is investigated in the three specific road environments.For the one-way road in the residential area,a supervisory assistant control system is proposed.The predictive model of vehicle motion in the system combines a vehicle kinematics model and a driver model.The driver model is obtained by fitting the driver data based on the potential field method,which reflects the driver's awareness of collision risks and driving habits and makes the prediction more reasonable in the long term.Based on the predictive model and the designed driving constraints,the constraint satisfaction problem is solved by the binary search,and the feedback control law is also presented to rectify the improper driving behaviors.The proposed supervisory assistant control system is tested and evaluated in the real vehicle experiment and driving simulator platform,respectively.The results show that the assistance of collision avoidance works well and the system has active influence on the driving behaviors.Furthermore,a formula of the probability of potential collision is proposed to estimate the collision risk when the vehicle is passing the intersection,which is verified by the Monte Carlo method.On the bend road in the suburbs,a safe steering control system is developed.This control system mainly contains a path following controller and a direct yaw moment controller.The former one plays a major role in the motion control of collision avoidance because it makes the vehicle track the bend road while avoiding the obstacles.The latter one is used to enhance the steering ability in the collision avoidance.Two kinds of model predictive control strategies are proposed for path planning and following.The first one is a hierarchical model predictive control strategy,where the upper layer replans the path of collision avoidance while the lower layer realizes the path following control and manages the constraints of the states of vehicle motion.The second one is a stochastic sample-based model predictive control strategy,which utilizes the inversed discrete cosine transform to generate the random control inputs,and it has low computational burden.This safe steering control system is tested based on the d SPACE platform.In the test,the controlled vehicle is able to avoid the obstacles on the bend road with low friction while keeping the motion stability.In the highway multi-agent environment,an intelligent hybrid control system is presented.It also has a hierarchical structure,where the upper layer is discrete coordination and the lower layer is continuous motion control.In the upper layer,a finite state machine is applied to identify the relative position of the surrounding vehicles and a fuzzy inference is employed to estimate the collision risk.As a result,the safe and reasonable driving mode is determined.In the lower layer,the vehicle lateral and longitudinal motion control is realized by linear model predictive control based on the linear tracking error model and inter-distance model.This intelligent hybrid control system is verified in a driver-in-the-loop platform.In the test,the controlled vehicle is able to respond to the collision risk from the surrounding vehicles and overtake by lane changing,then achieve the desired speed finally.