Research On Cooperative Lane-changing Strategy Of Vehicles Based On Game Theory

Driving safety is an important subject in the field of transportation.Autonomous vehicle is safer,efficient and can avoid many accidents caused by humans,thus providing a new idea for driving safety.However,if the autonomous vehicle lacks effective interaction mechanism in mixed traffic environment of both self-driving and human-driving,it cannot make full use of its advantages.Accidents might even happen as a result.Therefore,it has become the research focus to realize the vehicle cooperation between autonomous vehicles in mixed traffic environment of both human-driving and self-driving in VANET now.This paper is based on game theory.Aiming at one of the main behaviors in driving,lane-changing,we put forward a lane-changing strategy including a vehicle coordination mechanism: Lane-changing is regarded as a non-cooperative game behavior,vehicles select strategy for their own interests;To judge the need of lane-changing,the strategy considers both the scene of mandatory lane-changing and discretionary lane-changing,the vehicle seeks a better route at any time;According to the characteristics of lane-changing behaviors,the payoff function not only depends on the acceleration space which may be obtained under the premise of safety,but also on the influence which the following car has on the leading car;The strategy uses the payoff adjustment plan and assigns different weights to the vehicles which may affect the lane-changing for the payoff calculation through quantitative analysis of the vehicle’s driving style.In addition,in order to make the best use of road resources,the strategy encourages vehicles to compete with other vehicles for better traffic efficiency.In order to realize the cooperation between human-driving vehicles and autonomous vehicles,this paper designs a vehicle driving style computing system based on BP neural network: The system helps the autonomous vehicle have human-like abilities,so that the vehicle can understand and judge the human behaviors in the traffic scene;To analyze the vehicle driving style,both the rationaland irrationalfactors of driving behavior are considered,irrationalfactors cannot be ignored because they are more likely to cause accidents;The system not only takes the commonality of vehicles into account,but also takes the different driving styles of personality into account and thus can better adapt to the mixed driving scene.In addition,in order to analyze the irrationalfactors in driving behaviors better,this paper analyzes the factors that influence the competition and cooperation in the non-zero sum game from the view of psychological point.Finally,we verify the game theory based lane-changing strategy and driving style computing system through the simulation: The driving style computing system for various types of driving style evaluation and human evaluation of the coincidence rate of more than 90%,the system has the ability to understand human behavior and can provide the basis for cooperation between human-driving vehicles and autonomous vehicles;The game theory based lane-changing strategy in unmanned traffic environment or mixed traffic environment have shown the efficiency and stability.Combining with the driving style computing system,the lane-changing strategy can reduce the accidents and make the best effort to optimize the best efficiency on the premise of ensuring safety.