Optimizing Online Matching for Ride-Sourcing Services with Multi-Agent Deep Reinforcement Learning

TL;DR

This paper introduces a novel two-stage framework combining combinatorial optimization and multi-agent deep reinforcement learning to improve online matching efficiency in ride-sourcing services, reducing pickup times and enhancing system performance.

Contribution

It develops a new framework that optimally delays passenger matching and dynamically determines matching times using multi-agent deep reinforcement learning methods.

Findings

Significant reduction in average pickup distance and time.

Enhanced system efficiency demonstrated through extensive simulations.

Effective coordination of delayed matching with deep reinforcement learning.

Abstract

Ride-sourcing services are now reshaping the way people travel by effectively connecting drivers and passengers through mobile internets. Online matching between idle drivers and waiting passengers is one of the most key components in a ride-sourcing system. The average pickup distance or time is an important measurement of system efficiency since it affects both passengers' waiting time and drivers' utilization rate. It is naturally expected that a more effective bipartite matching (with smaller average pickup time) can be implemented if the platform accumulates more idle drivers and waiting passengers in the matching pool. A specific passenger request can also benefit from a delayed matching since he/she may be matched with closer idle drivers after waiting for a few seconds. Motivated by the potential benefits of delayed matching, this paper establishes a two-stage framework which incorporates a combinatorial optimization and multi-agent deep reinforcement learning methods. The multi-agent reinforcement learning methods are used to dynamically determine the delayed time for each passenger request (or the time at which each request enters the matching pool), while the combinatorial optimization conducts an optimal bipartite matching between idle drivers and waiting passengers in the matching pool. Two reinforcement learning methods, spatio-temporal multi-agent deep Q learning (ST-M-DQN) and spatio-temporal multi-agent actor-critic (ST-M-A2C) are developed. Through extensive empirical experiments with a well-designed simulator, we show that the proposed framework is able to remarkably improve system performances.