Dealing with elementary paths in the Kidney Exchange Problem

TL;DR

This paper introduces a new column generation approach for the kidney exchange problem that effectively handles chains of donation initiated by altruistic donors, overcoming NP-hardness in the pricing subproblem.

Contribution

It proposes a complete column generation scheme incorporating non-exact dynamic methods like NG-route relaxation and color coding heuristic for chains in kidney exchange.

Findings

Efficiently solves the pricing subproblem with chains of donation.

Develops non-exact dynamic approaches improving computational performance.

Addresses NP-hardness in the chain inclusion process.

Abstract

We study an elementary path problem which appears in the pricing step of a column generation scheme solving the kidney exchange problem. The latter aims at finding exchanges of donations in a pool of patients and donors of kidney transplantations. Informally, the problem is to determine a set of cycles and chains of limited length maximizing a medical benefit in a directed graph. The cycle formulation, a large-scale model of the problem restricted to cycles of donation, is efficiently solved via branch-and-price. When including chains of donation however, the pricing subproblem becomes NP-hard. This article proposes a new complete column generation scheme that takes into account these chains initiated by altruistic donors. The development of non-exact dynamic approaches for the pricing problem, the NG-route relaxation and the color coding heuristic, leads to an efficient column generation process.