Interaction Strategy For Automated Driving Considering The Risk Of Multi-vehicle Conflict

Automated driving technology is developing rapidly,in recent years.However,in some complex and dynamic driving situations,current automated driving technology still faces huge challenges.Especially in the case of automated driving and human driving sharing the road,it is of great significance to study the interaction strategy between automated vehicle and humandriven.This paper aims to solve the multi-vehicle interaction problem,and to proposes a global sorting-local gaming decision-making algorithm based on decision modeling.The aim is to propose an automated driving decision-making algorithm that can be applied in multiple scenarios,to improve the safety of interaction with other vehicles,and to enable automated driving to perform in harmony with human drivers.Firstly,the decision-making methods for automated multi-vehicle interaction decisions are reviewed and discussed.Existing methods can be divided into two categories: multi-vehicle collaboration and single-vehicle intelligence.On one hand,multi-vehicle collaboration methods tend to ignore the driver’s individual demands and the dynamic interaction between vehicles.On the other hand,although the methoda of single-vehicle intelligence can explain the interaction between vehicles,it is too difficult to consider the efficiency of the entire interaction system to solve traffic jams.Based on this,this paper combines the merits of multi-vehicle collaboration and single-vehicle intelligence,and proposes a global sorting-local gaming algorithm that considers both the entire system and a single driver.Secondly,the algorithm is extended and applied in a number of specific scenarios,including driving straight,turning left,changing lanes,and following a car,etc.The performance of the algorithm under different scenarios and different parameter designs is verificated through simulation.Results show that the algorithm can guarantee the safety and efficiency of interaction in all the multiple scenarios considered.On the other hand,by setting different initial conditions for simulation,results show that the algorithm can well consider factors such as driving style,intensity of conflict,decision time step and so on.Finally,a human-in-the-loop simulator experiments are designed.21 subjects were invited to participate in the test.The test scenarios,i.e.the three-vehicle interaction scenario include crossing and turning left at unsignalized intersections.Results show that when interacting with different human drivers,the algorithm can not only ensure sufficient safety,but also improve the traffic efficiency of the human drivers and the entire efficiency.In the experiments,the overall safety interaction rate of the algorithm is up to 97%.This research proposes a unified decision-making algorithm framework for multi-vehicle interaction of automated driving in complex scenarios.The algorithm combines the advantages of multi-vehicle collaboration and single-vehicle intelligence,and has potential application prospects.The reseach outcomes may work as a solid theoretical foundation for further decision algorithm developments and have many potential applicantions in real practice.