Solving the Team Orienteering Problem Using Decision Transformers

A. Guerrero Portoles, M. Escoto Gomar, A. A. Juan, Á. García Sánchez, W. Chen

The Team Orienteering Problem (TOP) is a well-established combinatorial optimization problem with applications in logistics, tourism, and other domains. Traditional solutions often rely on heuristics, metaheuristics, or exact algorithms, but deep learning-based approaches are emerging as promising alternatives. This study explores the use of Decision Transformers (DT)—a reinforcement learning framework based on sequence modeling—to generate high-quality solutions for the TOP. By utilizing a dataset of diverse routes with varying quality levels, the objective is for the DT to learn to predict effective action sequences that maximize the reward.

Palabras clave: Team Orienteering Problem (TOP), Combinatorial Optimization, Decision Transformers (DT), Reinforcement Learning

Métodos y aplicaciones de la IO II
13 de junio de 2025  09:00
Sala de prensa (MR 13)

J. Bautista Valhondo

M. Fernández de Dios, Á. M. González Rueda, J. R. Banga, J. González Díaz, D. R. Penas

I. Izco Berastegui, A. Al-Rahamneh, A. Serrano Hernandez, J. Faulin Fajardo