Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Regime-Calibrated Demand Priors for Ride-Hailing Fleet Dispatch and Repositioning

Indar Kumar, Akanksha Tiwari

Abstract

Effective ride-hailing dispatch requires anticipating demand patterns that vary substantially across time-of-day, day-of-week, season, and special events. We propose a regime-calibrated approach that (i) segments historical trip data into demand regimes, (ii) matches the current operating period to the most similar historical analogues via a similarity ensemble combining Kolmogorov-Smirnov distance, Wasserstein-1 distance, feature distance, variance ratio, event pattern similarity, and temporal proximity, and (iii) uses the resulting calibrated demand prior to drive both an LP-based fleet repositioning policy and batch dispatch with Hungarian matching. In ablation, a distributional-only metric subset achieves the strongest mean-wait reduction, while the full ensemble is retained as a robustness-oriented default that preserves calendar and event context. Evaluated on 5.2 million NYC TLC trips across 8 diverse scenarios (winter/summer, weekday/weekend/holiday, morning/evening/night) with 5 random seeds each, our method reduces mean rider wait times by 31.1% (bootstrap 95% CI: [26.5, 36.6]; Friedman chi-squared = 80.0, p = 4.25e-18; Cohen's d = 7.5-29.9). P95 wait drops 37.6% and the Gini coefficient of wait times improves from 0.441 to 0.409. The two contributions compose multiplicatively: calibration provides 16.9% reduction relative to the replay baseline; LP repositioning adds a further 15.5%. The approach requires no training, is deterministic and explainable, generalizes to Chicago (23.3% wait reduction using the NYC-built regime library without retraining), and is robust across fleet sizes (32-47% improvement for 0.5x-2.0x fleet scaling). Code is available at https://github.com/IndarKarhana/regime-calibrated-dispatch.

Regime-Calibrated Demand Priors for Ride-Hailing Fleet Dispatch and Repositioning

Abstract

Effective ride-hailing dispatch requires anticipating demand patterns that vary substantially across time-of-day, day-of-week, season, and special events. We propose a regime-calibrated approach that (i) segments historical trip data into demand regimes, (ii) matches the current operating period to the most similar historical analogues via a similarity ensemble combining Kolmogorov-Smirnov distance, Wasserstein-1 distance, feature distance, variance ratio, event pattern similarity, and temporal proximity, and (iii) uses the resulting calibrated demand prior to drive both an LP-based fleet repositioning policy and batch dispatch with Hungarian matching. In ablation, a distributional-only metric subset achieves the strongest mean-wait reduction, while the full ensemble is retained as a robustness-oriented default that preserves calendar and event context. Evaluated on 5.2 million NYC TLC trips across 8 diverse scenarios (winter/summer, weekday/weekend/holiday, morning/evening/night) with 5 random seeds each, our method reduces mean rider wait times by 31.1% (bootstrap 95% CI: [26.5, 36.6]; Friedman chi-squared = 80.0, p = 4.25e-18; Cohen's d = 7.5-29.9). P95 wait drops 37.6% and the Gini coefficient of wait times improves from 0.441 to 0.409. The two contributions compose multiplicatively: calibration provides 16.9% reduction relative to the replay baseline; LP repositioning adds a further 15.5%. The approach requires no training, is deterministic and explainable, generalizes to Chicago (23.3% wait reduction using the NYC-built regime library without retraining), and is robust across fleet sizes (32-47% improvement for 0.5x-2.0x fleet scaling). Code is available at https://github.com/IndarKarhana/regime-calibrated-dispatch.

Paper Structure

This paper contains 47 sections, 6 theorems, 8 equations, 10 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Contributions
  3. Related Work
  4. Ride-Hailing Dispatch and Matching
  5. Fleet Repositioning and Rebalancing
  6. Demand Forecasting for Dispatch
  7. Regime-Based Approaches
  8. Method
  9. Data Ingestion and Regime Library
  10. Similarity Ensemble
  11. Calibrated Demand Prior
  12. LP-Based Anticipatory Repositioning
  13. Batch Dispatch
  14. Theoretical Analysis
  15. LP Repositioning Optimality and Sensitivity
  16. ...and 32 more sections

Key Result

Theorem 1

The repositioning LP eq:lp_obj--eq:lp_budget is always feasible, bounded, and admits a polynomial-time optimal solution. Among all repositioning plans satisfying the supply, demand, linking, and budget constraints, the LP solution minimizes the weighted combination of repositioning cost and unserved

Figures (10)

  • Figure 1: System architecture. Historical TLC data is segmented into demand regimes; a six-metric similarity ensemble matches the current query block to historical analogues; the calibrated prior drives both LP repositioning and demand-aware dispatch.
  • Figure 2: Per-scenario wait reduction with 95% confidence intervals. All 8 scenarios show substantial improvement, with the largest gains on volatile demand periods (Friday night, New Year's).
  • Figure 3: Component decomposition: calibration and LP repositioning contribute additively to wait reduction. Bars show mean wait time; error bars show standard deviation across seeds.
  • Figure 4: Tail and fairness analysis. Left: CDF of wait times showing Cal+LP (solid) vs Replay (dashed). Right: Gini coefficient across scenarios---lower values indicate more equitable service.
  • Figure 5: Fleet sensitivity: wait reduction is robust across fleet scales ($0.5$--$2\times$). Calibration alone dominates at extreme scarcity; LP repositioning adds value when idle drivers are available.
  • ...and 5 more figures

Theorems & Definitions (10)

  • Theorem 1: LP Optimality and Feasibility
  • proof
  • Proposition 2: LP Sensitivity to Demand Estimation
  • proof
  • Corollary 3: Value of Calibration for Repositioning
  • Proposition 4: Calibration Error Bound via Chebyshev's Inequality
  • proof
  • Proposition 5: Similarity Metric Properties
  • proof
  • Lemma 6: Batch Matching Optimality