Research On Critical Scenarios Of Autonomous Vehicle Overtaking On Highway

The testing and verification of autonomous vehicles play a vital role in improving the safety of autonomous vehicles.Most autonomous driving companies and related research institutions use virtual testing methods to quickly and adequately test and verify autonomous vehicles,but there are currently few studies on complex critical scenarios under highway overtaking conditions.This paper studies the critical scenarios of overtaking conditions on highways for autonomous vehicles,in order to generate critical test cases for overtaking conditions on highways for closed road testing.The main research contents are as follows:(1)Based on the functional characteristics of autonomous vehicles,analyze the position and movement of the ego-vehicle and the interference vehicle,and then arrange and combine the relative positions and movement of the ego-vehicle and the interference vehicle based on the full combination strategy.The scenarios were screened based on the three principles of typicality,authenticity and simplicity,and a library of overtaking functional scenarios with test value was extracted.Finally,a two-way six-lane expressway section scene is selected as an example for scenario design,and the overtaking function scene of a two-way six-lane expressway section is obtained.(2)The overtaking behavior of autonomous vehicles on expressway was studied,and the overtaking process was decomposed into four sub-processes: overtaking preparation,left lane change,parallel overtaking and lane merging,and the safety factors affecting overtaking and the types of overtaking accidents were analyzed.On this basis,on the premise that the vehicle does not collide with other interfering vehicles,the minimum safe distance for overtaking on the expressway is deduced.Finally,a decision-making framework for overtaking based on hierarchical structure is proposed.(3)Based on Model Predictive Control(MPC),a horizontal and vertical decoupling autonomous vehicle motion planning and control algorithm based on vertical safety priority is selected.A three-degree-of-freedom single-track dynamic model that is more mature and highly adaptable is selected,and idealized assumptions are made.Then the nonlinear model predictive control problem is transformed into a linear time-varying model predictive control problem,and further transformed into a quadratic programming problem.Then according to the control objective,the corresponding objective function is established,and corresponding constraints are added to the objective function.Finally,based on the Matlab/Simulink and Carsim co-simulation platform,a simulation analysis of some overtaking conditions is carried out to verify the feasibility and effectiveness of the overtaking algorithm.(4)Based on the overtaking function scenario,design the dynamic scene parameters of the main vehicle and each interfering vehicle,and use the three-parameter combination to combine the dynamic scene parameters.Then,based on the simulation method,the dynamic scene parameters after the preliminary screening are further screened using collision time,corner distance and longitudinal maximum deceleration index,and 70 groups of key scenes under highway overtaking conditions are obtained.Aiming at the problem of repetitive and similar scenes in dynamic scenes,a clustering method based on weighted Euclidean distance is used to cluster the selected key scenes,and 11 groups of key scenes for overtaking on highway are obtained.Finally,the key scenes were deconstructed,and the deconstructed information was recombined to generate 88 sets of key test cases for autonomous vehicle highway overtaking conditions for closed road site testing.