Overview
The research introduces HCRide, a Human-Centered Ride-hailing system designed to integrate passenger fairness and driver preference into order dispatch systems. This system aims to address potential conflicts between these objectives while maintaining overall system efficiency. HCRide is based on a novel multi-agent reinforcement learning algorithm named Harmonization-oriented Actor-Bi-Critic (Habic).
Research Context
Order dispatch systems are integral to ride-hailing services, impacting operator revenue, driver profit, and passenger experience. Existing work predominantly prioritizes system efficiency, often measured by operator revenue. This focus can result in suboptimal experiences for both passengers and drivers. The challenge lies in balancing passenger fairness and driver preference, as optimizing one can potentially compromise the other, without sacrificing overall system efficiency.
Approach
The HCRide system incorporates the Harmonization-oriented Actor-Bi-Critic (Habic) algorithm. Habic comprises three main components:
- A multi-agent competition mechanism.
- A dynamic Actor network.
- A Bi-Critic network.
These components are designed to optimize system efficiency and passenger fairness while considering driver preference. The evaluation of HCRide involved extensive testing using two real-world ride-hailing datasets obtained from Shenzhen and New York City.
Findings
Experimental results from the evaluation demonstrated improvements across several key metrics:
- HCRide effectively improved system efficiency by 2.02% when compared to state-of-the-art baselines.
- Passenger fairness was enhanced by 5.39% relative to existing baselines.
- Driver preference saw an improvement of 10.21% compared to state-of-the-art baselines.
Why This Matters
The development of HCRide addresses the critical need for ride-hailing dispatch systems that move beyond solely optimizing operator revenue. By integrating passenger fairness and driver preference without compromising efficiency, it offers a framework for more balanced and sustainable ride-hailing operations.