Online car-hailing order allocation method suitable for passenger waiting time distribution changing along with time

Abstract

The invention discloses an online car-hailing order allocation method suitable for passenger waiting time distribution changing with time. The method comprises the steps of calculating an optimal matching interval corresponding to a batch processing algorithm in a distribution change range according to the passenger waiting time distribution change range, and training different deep reinforcementlearning models for different passenger waiting time distributions, combining the batch processing algorithms with different matching intervals and the deep reinforcement learning model adapting to different passenger waiting time distributions into an algorithm set; when an order arrives, adopting a preset selection model to select one algorithm from the algorithm set to allocate the order, and updating parameters in the selection model according to income feedback; and detecting whether passenger waiting time distribution is changed or not in combination with income feedback, and if so, restarting the selection model and emptying a feedback record. Distribution can be continuously adjusted along with the change of waiting time and number distribution of passengers, thereby maximizing theincome of an online car-hailing platform.

Description

technical field [0001] The present invention relates to the technical field of dispatching and planning, in particular to an online car-hailing order allocation method that is suitable for passenger waiting time distribution changing with time. Background technique [0002] With the rapid development of the economy, the lack of autonomy and personalization of public transportation has become increasingly prominent. When the existing public transportation cannot meet everyone's needs, the emergence of online car-hailing platforms meets people's personalized travel needs and fills in the shortcomings of public transportation. How to efficiently allocate orders to online car-hailing vehicles not only affects the experience of passengers, but also affects the revenue of online car-hailing platforms. [0003] The online car-hailing order assignment problem can be regarded as an online bipartite matching problem: waiting passengers belong to nodes on one side of the bipartite gra...

AI Technical Summary

Problems solved by technology

At the same time, compared with the quantity distribution, the waiting time distribution is always difficult to directly count because the remaining waiting time of the matched passengers is not known.

Therefore, it is difficult to adjust the matching strategy by directly observing or counting long-term historical data to predict changes in waiting time distribution.

[0005] Since the distribution of waiting time is difficult to be directly counted, the traditional matching strategy always does not consider the change of waiting time distribution.

In other words, the traditional order matching strategy always performs better under certain distributions but cannot maintain the original efficiency when the distribution changes

On the other hand, the traditional matching strategy that is difficult to adapt to the changing distribution also greatly affects the passenger experience

Images