Iterated Granular Neighborhood Algorithm for the Taxi Sharing Problem

Abstract

One of the most popular issues that we can find in cities is transportation problems: traffic jams, pollution and the transportation cost fees. The concept of taxi sharing is considered as a promising idea to reduce some of the transportation problems. A group of people travels from the same origin to different destinations. Our goal is to assign them to several taxis while reducing the cost of all trips. The taxi sharing problem is NP-hard, since it is a variant of the car pooling problem. We adapt Capacitated Vehicle Routing Problem (CVRP) to solve the taxi sharing problem, in which goods are changed by passengers and trucks by taxis. We describe a new algorithm, called Iterated Granular Neighborhood Algorithm (IGNA), based on the use of the restricted swap neighborhoods in the local search phase, eliminating moves that involve long arcs that may not be part of the best solution. We empirically analyze our algorithm solving different real-like instances of the problem with 9 to 57 passengers. The results show that the proposed IGNA is quite competitive with the parallel micro evolutionary algorithm (p$\mu$EA).

This research is partially funded by the Spanish Ministry of Economy and Competitiveness and FEDER under contract TIN2017-88213-R (6city); and Universidad de Málaga, Andalucía Tech, Consejería de Economía y Conocimiento de la Junta de Andaluía, and European Regional Development Fund under grant number UMA18-FEDERJA-003 (PRECOG).