Research On Multi-objective Coordinated Control Strategies For Vehicle Collision-avoidance In Highway Driving

In emergency collision avoidance conditions,most accidents occur due to human factors such as delays in response,misjudgments,and improper operations.Hence,the advanced driver assistance system(ADAS)has played an increasing role in traffic accidents,through enhancing the ability of driver's perception and decision-making.Particularly,collision avoidance system(CAS),as a main supplement to autonomous driving technology,has great potential in reducing traffic accidents caused by driver's limitations.However,CAS technology in past years mainly focused on emergency collision-avoidance condtions at lower speeds by braking maneuver.In addition,collision avoidance by steering maneuver in the existing CAS products still depended on the driver's handling ability;therefore,the collision risk caused by human factors also remains.On the other hand,in case of emergency driving situations at high speeds,due to the highway driving and sudden changes in traffic environment,the collision manoeuvre should be apllied in a very short duration;as a result,the real-time performance of the collision algorithm has become one of the key issues for developing CAS technology.More specifically,when the vehicle performs an emergency evasive manoeuvre in highway driving,the vehicle dynamics exhibit nonlinear characteristics,which the cause of vehicle instability.The traditional electronic stability control system can enhance vehicle stability in critical situations,but the actions for vehicle stabilisation also conflict with the objective of collision avoidance.To this end,this paper will focus on the above problems and study the highly autonomous collision-avoidance system in highway driving,to accommodate both a low computational complexity and an effective performance of collision avoidance and vehicle stabilisation.The main research in this paper are summarized as follows:(1)In order to explore the the dynamics characteristics of vehicles and validate the effectiveness of the proposed control method,the nonlinear vehicle model was developed.In addition,an unscented Kalman filter(UKF)employing the established vehicle model was designed in this paper.The designed estimator is used to obtain vehicle states that are not available from the sensors deployed in the vehicle,such as sideslip angle and tire slip angles.Then the observer can provide the relevant vehicle states to the proposed collision-avoidance controller to improve the vehicle stafety.(2)This paper explores the intervention criterion for the CAS in emergency collision avoidance situations at high speeds.Initially,based on situation assessment,a rule-based decision logic system is modeled by finite state machine(FSM)technology.The risk monitor is designed to use the appropriate risk metrics with respect to time to collision(TTC),Braking-Threat-Number(BTN)and Steering-Threat-Number(STN);furthermore,the metrics decide on a suitable intervention manoeuvre in a hazard situati on.In addition,with consideration of the emergency situation related to surrounding vehicles,this paper developed the appropriate decision-rules,which corresponds to the location of surrounding vehicles in an adjacent lane.Particularly,when surroundi ng vehicles driving in an adjacent lane with overlapping subareas,a collision risk “vehicle-road” model was developed,taking into account risk related to road adhesion,stabilization performance of vehicle,and constraints of vehicle stafety.Then based on the collision risk model and predictions of vehicles,the collision decision was evaluated over the prediction horizion.(3)To improve the real-time performance of the collision avoidance system,this paper proposed a new collsion avoidance strategy t hat the process of collision avoidance is divided into two stages: in the “steer-in” stage,the autonomous steering guides the subject vehicle to avoid the collision with a constant centripetal acceleration.After the subject vehicle has been out of the co llision with the obstacle according to the collision avoidance constraints,the second stage begins,namely guided along the centerline of the adjacent lane.In this stage,a model predictive control(MPC)-based algorithm is used to track the centerline of the lane.To achive the multi-objective requriments for vehicle stabilization and road safety,the reference yaw rate was modified in this paper,and based on that,we introduce the differential braking controller with a comprehensive treatment to the multi-objective requriments.Addationally,accounting for safe gaps to surrounding vehicles in an adjacent lane,the londitudinal motion planner is designed by a model-predictive control algorithm,finally the control allocation with propusion was developed,and the weighted least squares method was used to distribute the tire longitudinal force.(4)To evaluate the effectiveness of the proposed collision avoidance system,a series of model-in-the-loop(MIL)simulations,hardware-in-the-loop(HIL)and vehicle-in-the-loop(HIL)experiments are conducted in various scenarios such as various vehicle speeds,road adhesion,and surrounding-vehicle settings.The results have successfully demonstrated the real-time performance of the collision avoidance algorithm.In addition,the results validate that with the assitance of the collision avoidance system,the vehicle can effectively achive better balance between collision avoidance and vehicle yaw stabilization at high speeds.