Research On Vehicle Competition Game And Autonomous Coordination Allocation Strategy For Shared Travel

With the increase in residents’ demand for travel,online car-hailing and taxis have become essential means of transportation in people’s daily lives and occupy an important position.Nowadays,with the integration of the "Internet+" traditional transportation industry,the order scheduling network platform has emerged.The existing shared vehicle scheduling methods are mainly based on the travel platform for centralized order dispatch,aiming to improve the global order matching rate.However,there are two problems with this centralized scheduling method: 1)The platform sacrifices personalization in order to ensure the fairness of drivers’ orders.For example: when the vehicle is close to the order,the vehicle cannot independently choose to go to receive the order but must be assigned by the platform,resulting in the dilemma that nearby vehicles cannot receive the order.2)When the number of vehicles to be dispatched increases,it will cause scheduling bottlenecks due to the enormous amount of computation on the platform,reducing efficiency.Although the platform does not restrict the vehicle cruising mode represented by taxis,it can independently receive orders.However,due to the lack of global order information and information on the selection of surrounding vehicles,drivers fall into a vicious competition,wasting transport capacity and resources.For example,when the number of vehicles in the area is greater than the number of orders,multiple drivers take orders out of price "temptation," and conflicts arise in order to grab.In this scenario,time costs and fuel consumption are spent on drivers who do not take orders,while from a global perspective,it consumes capacity resources and reduces the order response rate.To address the above problems,we propose a distributed autonomous vehicle coordination order-receiving model,which not only ensures the individualization of the allocation process but also largely avoids vicious competition.Specific research contents are as follows:1)The game theory is used to model the scene of vehicle order grabbing in the vehicle allocation problem.This paper assumes that the premise of vehicle order grab behavior is:(1)All vehicles grab orders simultaneously.(2)In the game area,all vehicles can know the status,order receiving the action,and income of other vehicles.Based on this,this paper proposes a distributed vehicle confrontation game model based on complete information and a coordinated allocation strategy based on Nash equilibrium and gives a theoretical method for solving the optimal solution according to different game scenarios.2)Optimizing the vehicle’s autonomous coordination and allocation strategy improves the solution efficiency of order distribution.This paper proposes a vehicle coordination allocation strategy MANE(Multi-Agent Nash Equilibrium)based on a reinforcement learning framework.Specifically,by modeling and analyzing the competitive relationship between drivers,the Nash equilibrium solution method is combined with the reinforcement learning strategy to avoid vehicle order conflicts and increase the cumulative income of drivers.3)A simulator was designed and developed to simulate the autonomous vehicle coordination order receiving mode in a natural traffic environment,and the MANE was verified using the travel order data of a domestic platform.The experimental results show that using the MANE strategy increases the driver’s average income by 8.1% compared with the baseline algorithm,and the order response rate is increased by 9.47%.The average vehicle utilization rate increased by 10.53%,and the algorithm convergence speed increased by about 33%.