TAXI2: Joint Pairing and Vehicle Assignment for Two-Passenger Shared Taxis

Afzaal Hassan, Irene Moser, Mark Wallace, Daniel Harabor

Keywords: Shared taxi problem, Dynamic ride matching, Integer linear programming (ILP)

TL;DR: The paper introduces TAXI2, an ILP that jointly chooses rider pairs and assigns vehicles (or solo trips) in one shot to minimise vehicle hours travelled, with both flexible (small delays allowed) and strict (no delays) deadline variants.

Abstract: We study the two-passenger shared taxi problem in which a dedicated fleet must serve dynamic requests under spatio-temporal constraints. We present TAXI2, an integer linear program (ILP) based framework that, in a single optimisation, forms feasible rider pairs and assigns vehicles to pairs or solo trips, with the dual objective of minimising vehicle hours travelled (VHT) and maximising service rate. Using one hour of New York City demand restricted to Manhattan (about 24000 requests), we show that optimal solutions are tractable in rolling horizons when modest detours are allowed. We study both relaxed-deadline and strict-deadline variants and observe that enforcing deadlines improves reliability but increases computational effort. To reduce solve times, we introduce \emph{Octo-Prune}, which enumerates feasible matches and retains eight high-quality candidates per rider. Octo-Prune achieves near-optimal VHT and service rates in our experiments while substantially reducing runtime, indicating practical viability for large-scale deployment.

Area: Search, Optimization, Planning, and Scheduling (SOPS)

Generative A I: I acknowledge that I have read and will follow this policy.

Submission Number: 1600