Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LEAD: Towards Learning-Based Equity-Aware Decarbonization in Ridesharing Platforms

Mahsa Sahebdel, Ali Zeynali, Noman Bashir, Prashant Shenoy, Mohammad Hajiesmaili

TL;DR

LEAD presents a learning-based, equity-aware framework for decarbonizing ridesharing by balancing system-wide emissions with fair driver utilities while preserving rider wait times. It formulates ride assignment as a forward-looking bipartite matching problem and uses TD learning to estimate future deadhead and trip distances, feeding an ILP that jointly minimizes total emissions and fairness across drivers, with batch-based decisions controlled by parameter $\eta$. On RideAustin data, LEAD achieves substantial emissions reductions (up to $56.6\%$ vs the fairness-focused baseline) and improves driver fairness, with performance benefiting from increased low-emission vehicle penetration and modest sensitivity to batch duration; rider wait times remain competitive relative to baselines. These results demonstrate that incorporating temporal dependencies and equity considerations yields more sustainable and equitable ridesharing, providing practical guidance for deployment and future extensions to rider-focused fairness and short-term driver utility dynamics.

Abstract

Ridesharing platforms such as Uber, Lyft, and DiDi have grown in popularity due to their on-demand availability, ease of use, and commute cost reductions, among other benefits. However, not all ridesharing promises have panned out. Recent studies demonstrate that the expected drop in traffic congestion and reduction in greenhouse gas (GHG) emissions have not materialized. This is primarily due to the substantial distances traveled by the ridesharing vehicles without passengers between rides, known as deadhead miles. Recent work has focused on reducing the impact of deadhead miles while considering additional metrics such as rider waiting time, GHG emissions from deadhead miles, or driver earnings. However, most prior studies consider these environmental and equity-based metrics individually despite them being interrelated. In this paper, we propose a Learning-based Equity-Aware Decarabonization approach, LEAD, for ridesharing platforms. LEAD targets minimizing emissions while ensuring that the driver's utility, defined as the difference between the trip distance and the deadhead miles, is fairly distributed. LEAD uses reinforcement learning to match riders with drivers based on the expected future utility of drivers and the expected carbon emissions of the platform without increasing the rider waiting times. Extensive experiments based on a real-world ridesharing dataset show that LEAD improves the defined notion of fairness by 150% when compared to emission-aware ride-assignment and reduces emissions by 14.6% while ensuring fairness within 28--52% of the fairness-focused baseline. It also reduces the rider wait time, by at least 32.1%, compared to a fairness-focused baseline.

LEAD: Towards Learning-Based Equity-Aware Decarbonization in Ridesharing Platforms

TL;DR

LEAD presents a learning-based, equity-aware framework for decarbonizing ridesharing by balancing system-wide emissions with fair driver utilities while preserving rider wait times. It formulates ride assignment as a forward-looking bipartite matching problem and uses TD learning to estimate future deadhead and trip distances, feeding an ILP that jointly minimizes total emissions and fairness across drivers, with batch-based decisions controlled by parameter . On RideAustin data, LEAD achieves substantial emissions reductions (up to vs the fairness-focused baseline) and improves driver fairness, with performance benefiting from increased low-emission vehicle penetration and modest sensitivity to batch duration; rider wait times remain competitive relative to baselines. These results demonstrate that incorporating temporal dependencies and equity considerations yields more sustainable and equitable ridesharing, providing practical guidance for deployment and future extensions to rider-focused fairness and short-term driver utility dynamics.

Abstract

Ridesharing platforms such as Uber, Lyft, and DiDi have grown in popularity due to their on-demand availability, ease of use, and commute cost reductions, among other benefits. However, not all ridesharing promises have panned out. Recent studies demonstrate that the expected drop in traffic congestion and reduction in greenhouse gas (GHG) emissions have not materialized. This is primarily due to the substantial distances traveled by the ridesharing vehicles without passengers between rides, known as deadhead miles. Recent work has focused on reducing the impact of deadhead miles while considering additional metrics such as rider waiting time, GHG emissions from deadhead miles, or driver earnings. However, most prior studies consider these environmental and equity-based metrics individually despite them being interrelated. In this paper, we propose a Learning-based Equity-Aware Decarabonization approach, LEAD, for ridesharing platforms. LEAD targets minimizing emissions while ensuring that the driver's utility, defined as the difference between the trip distance and the deadhead miles, is fairly distributed. LEAD uses reinforcement learning to match riders with drivers based on the expected future utility of drivers and the expected carbon emissions of the platform without increasing the rider waiting times. Extensive experiments based on a real-world ridesharing dataset show that LEAD improves the defined notion of fairness by 150% when compared to emission-aware ride-assignment and reduces emissions by 14.6% while ensuring fairness within 28--52% of the fairness-focused baseline. It also reduces the rider wait time, by at least 32.1%, compared to a fairness-focused baseline.
Paper Structure (13 sections, 9 equations, 5 figures, 1 algorithm)

This paper contains 13 sections, 9 equations, 5 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. The LEAD Algorithm
  5. Learning Expected Emissions and Utility
  6. Batched Emission-Aware Fair Assignment
  7. Experimental Evaluation
  8. Experimental Setup
  9. Impact on Per-Trip Emissions
  10. Impact on Driver's Fairness
  11. Impact on Rider's Wait Time
  12. Conclusion
  13. Acknowledgements

Figures (5)

  • Figure 1: Overview of LEAD. During each batch, LEAD evaluates the long-term emissions and utilities and uses them to construct the weighting of the ILP problem of \ref{['eq:LEAD_ILP']}. After solving the optimization problem and matching ride requests with drivers, LEAD updates the value functions based on the deadhead and trip distances of the served requests in the batch.
  • Figure 2: Emissions reduction performance: (a) emission per trip for different algorithms as a function of batch duration, (b) impact of the percentage of low emission vehicles (LEVs) in the fleet on LEAD for different batch durations and $\eta$ = 5, and (c) the impact of increasing emissions for an increase in fairness, captured using $\eta$, for a batch duration of 5 minutes. Here, $\eta$ specifies extra emissions that the algorithm incurs to reduce unfairness by 1km.
  • Figure 3: Fairness performance: (a) normalized fairness for different algorithms as a function of batch duration, (b) impact of the percentage of low emission vehicles (LEVs) in the fleet on LEAD for different batch durations and $\eta$ = 5, and (c) the impact of increasing emissions for an increase in fairness, captured using $\eta$, for a batch duration of 5 minutes. Here, $\eta$ specifies extra emissions that the algorithm incurs to reduce unfairness by 1km.
  • Figure 4: Wait time performance: (a) wait time for different algorithms as a function of batch duration, (b) impact of the percentage of low emission vehicles (LEVs) in the fleet on LEAD performance for different batch durations and $\eta$ = 5, and (c) the impact of increasing emissions for an increase in fairness, captured using $\eta$, for a batch duration of 5 minutes. Here, $\eta$ specifies extra emissions that the algorithm incurs to reduce unfairness by 1km.
  • Figure : LEAD($\mathcal{R}_b$, $\mathcal{V}_b$)