A branch-cut-and-price algorithm for a dial-a-ride problem with minimum disease-transmission risk

TL;DR

This paper introduces a risk-aware dial-a-ride problem model that minimizes travel cost and disease transmission risk, solved efficiently with a novel branch-cut-and-price algorithm, demonstrated on real-world datasets.

Contribution

It develops a new risk-aware formulation of DARP with a min-max trip-based model and an exact BCP solution method, including a novel labeling algorithm for risk minimization.

Findings

Optimal solutions for small and medium instances within 5 minutes.

Large instances solved optimally within one hour for most cases.

The approach effectively balances travel cost and disease transmission risk.

Abstract

This paper investigates a variant of the dial-a-ride problem (DARP), namely Risk-aware DARP (RDARP). Our RDARP extends the DARP by (1) minimizing a weighted sum of travel cost and disease-transmission risk exposure for onboard passengers and (2) introducing a maximum cumulative exposure risk constraint for each vehicle to ensure a trip with lower external exposure. Both extensions require that the risk exposure of each onboard passenger is propagated in a minimized fashion while satisfying other existing constraints of the DARP. To fully describe the RDARP, we first provide a three-index arc-based formulation and reformulate the RDARP into an equivalent min-max trip-based formulation, which is solved by an exact Branch-Cut-and-Price (BCP) algorithm. For each node in the branching tree, we adopt the column generation method by decomposing the problem into a master problem and a subproblem. The latter takes the form of an elementary shortest path problem with resource constraints and minimized maximum risk (ESPPRCMMR). To solve the ESPPRCMMR efficiently, we develop a new labeling algorithm and establish families of resource extension functions in compliance with the risk-related resources. We adopt a real-world paratransit trip dataset to generate the RDARP instances ranging from 17 to 55 heterogeneous passengers with 3 to 13 vehicles during the morning and afternoon periods. Computational results show that our BCP algorithm can optimally solve all small- and medium-sized instances (32 or fewer passengers) within 5 minutes. For the large instances (39 to 55 passengers), our BCP algorithm can solve 23 of 30 instances optimally within a time limit of one hour.