Research On Active Collision Avoidance Based On Modeling On Critical Collision Scenarios

Active collision avoidance,as one of the most important functions of intelligent driving assistance systems or autonomous driving systems,directly affects the safety of cars and the level of vehicle intelligence.It is extremely important in improving driving safety and reducing the frequency of collision accidents.The modeling of critical collision scenarios is also of great significance for describing critical driving scenarios on structured roads,establishing standardized,normalized descriptions of collision hazards under dangerous conditions,or for generating subsequent critical collision scenarios.The main work done in this paper is as follows:(1)For the existing collision hazard evaluation methods,it is difficult to evaluate different hazard scenarios horizontally,and the numerical intervals vary widely,the evaluation methods are complicated,and there is no normalized description form.Collision hazards in longitudinal,lateral,and parallel working conditions are analyzed and physical modeling based on the concepts of microscopic motion processes and headway,and the hazard description formulas corresponding to the working conditions are derived to form a dimensionless and normalized comprehensive description and evaluation method.(2)There is no in-depth study on the description and evaluation form of the dangerous collision mechanism and the collision hazard in the active collision avoidance decision-making algorithm at present.The single form of hazard description based on time headway or kinematics leads to problems such as inconsistent evaluation methods and incomparable numerical values.Therefore,it is difficult to quantitatively evaluate the functions of different intelligent driving systems.It also leads to inadequate standardization of risk assessment modeling in function development.Active collision avoidance also has problems such as single mode and poor performance of real vehicles.This article uses the aforementioned modeling theory of dangerous collision scenarios and combines the advantages of lateral steering obstacle avoidance under low friction conditions to form active collision avoidance decisions.(3)Aiming at the problem that the driving mode and braking mode oscillate around the sliding curve in the existing control,a mode arbitration strategy based on the dead zone of the sliding curve is proposed,and combine the modeling of longitudinal inverse dynamics and lateral dynamics to form a lower-level controller for collision avoidance system.(4)Considering deceleration constraints and impact constraints during longitudinal control,combined with the accuracy and real-time requirements of real vehicle control,model predictive control theory is used to form the upper vertical control strategy;Lateral collision avoidance control requires trajectory planning under the premise of considering the trajectory planning constraints and the road driving environment.The accurate tracking of the ideal lateral trajectory relies on the method of least squares estimation of the path curvature based on arc fitting and the Ackerman steering geometric model considering the stability factor at high speeds.Feedforward and optimal LQR control are used for lateral control.Based on PanoSim / Simulink,different simulation conditions are designed to illustrate the effectiveness of the comprehensive collision hazard description method.At the same time,the active collision avoidance system can select different working modes according to the driving environment and decision strategies to achieve the expected working effect and control accuracy,and improve driving stability and security.