Vehicle sharing service order dispatching method and system based on reinforcement learning

Abstract

The invention provides a vehicle sharing service order dispatching method and system based on reinforcement learning, and the method comprises the steps of collecting the information of passengers andthe information of vehicles, carrying out the corresponding limitation according to the demands of the passengers and the start and end points of the passengers, and searching vehicles meeting the space-time constraint of the passengers to obtain a candidate vehicle set; calculating dispatching factors between the vehicles in all candidate vehicle sets and the current passenger, wherein the dispatching factors comprise the detour ratio of the vehicles, the seat utilization rate, the hidden income of the vehicles and the future income; arranging all the candidate vehicles in an ascending orderaccording to the detouring ratio of the vehicles, and selecting a final candidate vehicle set; inputting the dispatching factor of each vehicle into the deep evaluation network for evaluation, selecting the vehicle with the optimal evaluation result to return to passengers and vehicles, ending the current evaluation if the training of the deep evaluation network is completed, otherwise, trainingthe deep evaluation network in combination with a reinforcement learning strategy and a gradient descent method, and supporting the next evaluation by using a new deep evaluation network.

Description

technical field [0001] The invention relates to the field of vehicle sharing, in particular to a method and system for dispatching vehicle sharing service orders based on reinforcement learning. Background technique [0002] Ride-sharing refers to a travel method in which several people on the same route take the same vehicle to carry out certain activities, and the fare is equally shared by passengers. With the development of cities and the popularity of sharing models, urban congestion and traffic pollution have become important issues affecting urban development. In order to solve these problems, vehicle sharing has become one of the main ways for citizens to travel. There are many methods of vehicle sharing. In these methods, the main input includes two parts, namely, the set of passengers and the set of vehicles. Then the vehicle sharing method finds the vehicle that satisfies the space-time constraints of the passengers according to the requirements of the passengers, ...

AI Technical Summary

Problems solved by technology

[0003] Massobrio et al. proposed a genetic algorithm to solve the problem of dispatching vehicles with multiple destinations at the same starting point and considered the waiting time of passengers and the overall driving distance of the vehicle at the same time. Sharing can only be done at the same location. In many cases, the location of passengers does not meet the constraints

Vinicius et al. proposed an activity-oriented carpooling method. In the activity-oriented carpooling method, the destination of passengers is not fixed, that is, passengers only need to reach the destination where they can complete their activities. However, in this method, the vehicle The detour distance is not specifically considered. Zhu et al. proposed a method to reduce the amount of calculation with QoS constraints. These constraints include passenger waiting time, vehicle travel distance, and detour ratio. did not plan

Images