Research On Risk Estimation And Avoidance At Unsignalized Intersection Under Vehicle-road Coordination Environment

The intersection generally takes the attribute of traffic congestion and vehicle accidents,which seriously affects traffic efficiency in the road network.How to solve the problem of cooperative optimization in unsignalized intersections has become the key in the field of intelligent transportation.It is also an important part to improve the vehicle-road cooperative automatic driving system in complex traffic environments.This thesis takes the vehicle-road cooperative environment as the background,connected autonomous vehicle as research object,driving safety field and model predictive control as the theoretical basis,improving traffic efficiency as well as ensuring safety as the aims,and research on cooperative decision-making for autonomous vehicles at unsignalized intersection.Under the framework of driving safety field,an extended driving risk estimation method based on conflict relationship and minimum safety distance is proposed.The new risk estimation method is integrated into the risk-avoidance decision-making optimization,and a decision-making algorithm that comprehensively considers multiple goals is designed.The specific research content of this paper mainly includes three parts:(1)Based on the driving safety field,an extended driving risk estimation model(E-DRE)is proposed aiming at driving risk assessment at unsignalized intersection.Considering the global driving direction of vehicles and their conflict relationship,the virtual lane and the minimum safety distance is used to design the collision risk between conflicting vehicles.The risk estimation model is able to improve the perception sensitivity of vehicle collisions.(2)Aiming at the problem of risk-avoidance decision-making at unsignalized intersections under vehicle-road coordination,a multi-objective risk avoidance model predictive control(MO-RAMPC)algorithm is designed.Based on the MPC framework and E-DRE model,the MO-RAMPC algorithm comprehensively considers safety,comfort,flexibility and environmental characteristics.The cost function and constraints are designed for the risk-avoidance demand,and finally the problem is discretized and transformed into a quadratic programming problem to solve.(3)Based on Matlab,collision scenarios are designed in a simulated intersection to verify the logic of MO-RAMPC and its risk-avoidance performance.The MO-RAMPC algorithm is tested and evaluated from the perspective of security and rapidity.By optimizing cost function,the collision risk between conflicting vehicles is reduced,and the driving safety and environmental performance are at an optimal level,providing a risk-avoidance decision-making scheme for vehicles approaching the intersection.With the extended risk estimation model,the algorithm can realize the cooperation of perception and risk avoidance.Based on CARLA,an unsignalized intersection scenario is constructed to verify the effectiveness of E-DRE method.Results show that,compared with the original estimation model,the E-DRE method is more sensitive to the collision risk,and the average time to detect risk is advanced by 1.7 seconds.The experimental results of risk avoidance based on numerical simulation in Matlab show that MO-RAMPC can complete the task of avoiding danger in single-vehicle and multi-vehicle risk scenarios to a certain extent,and improve the vehicle rapidity.