-
A New Constructive Heuristic driven by Machine Learning for the Traveling Salesman Problem
Authors:
Umberto Junior Mele,
Luca Maria Gambardella,
Roberto Montemanni
Abstract:
Recent systems applying Machine Learning (ML) to solve the Traveling Salesman Problem (TSP) exhibit issues when they try to scale up to real case scenarios with several hundred vertices. The use of Candidate Lists (CLs) has been brought up to cope with the issues. The procedure allows to restrict the search space during solution creation, consequently reducing the solver computational burden. So f…
▽ More
Recent systems applying Machine Learning (ML) to solve the Traveling Salesman Problem (TSP) exhibit issues when they try to scale up to real case scenarios with several hundred vertices. The use of Candidate Lists (CLs) has been brought up to cope with the issues. The procedure allows to restrict the search space during solution creation, consequently reducing the solver computational burden. So far, ML were engaged to create CLs and values on the edges of these CLs expressing ML preferences at solution insertion. Although promising, these systems do not clearly restrict what the ML learns and does to create solutions, bringing with them some generalization issues. Therefore, motivated by exploratory and statistical studies, in this work we instead use a machine learning model to confirm the addition in the solution just for high probable edges. CLs of the high probable edge are employed as input, and the ML is in charge of distinguishing cases where such edges are in the optimal solution from those where they are not. . This strategy enables a better generalization and creates an efficient balance between machine learning and searching techniques. Our ML-Constructive heuristic is trained on small instances. Then, it is able to produce solutions, without losing quality, to large problems as well. We compare our results with classic constructive heuristics, showing good performances for TSPLIB instances up to 1748 cities. Although our heuristic exhibits an expensive constant time operation, we proved that the computational complexity in worst-case scenario, for the solution construction after training, is $O(n^2 \log n^2)$, being $n$ the number of vertices in the TSP instance.
△ Less
Submitted 17 August, 2021;
originally announced August 2021.
-
Machine Learning Constructives and Local Searches for the Travelling Salesman Problem
Authors:
Tommaso Vitali,
Umberto Junior Mele,
Luca Maria Gambardella,
Roberto Montemanni
Abstract:
The ML-Constructive heuristic is a recently presented method and the first hybrid method capable of scaling up to real scale traveling salesman problems. It combines machine learning techniques and classic optimization techniques. In this paper we present improvements to the computational weight of the original deep learning model. In addition, as simpler models reduce the execution time, the poss…
▽ More
The ML-Constructive heuristic is a recently presented method and the first hybrid method capable of scaling up to real scale traveling salesman problems. It combines machine learning techniques and classic optimization techniques. In this paper we present improvements to the computational weight of the original deep learning model. In addition, as simpler models reduce the execution time, the possibility of adding a local-search phase is explored to further improve performance. Experimental results corroborate the quality of the proposed improvements.
△ Less
Submitted 2 August, 2021;
originally announced August 2021.
-
Benchmark Instances and Optimal Solutions for the Traveling Salesman Problem with Drone
Authors:
Mauro Dell'Amico,
Roberto Montemanni,
Stefano Novellani
Abstract:
The use of drones in logistics is gaining more and more interest, and drones are becoming a more viable and common way of distributing parcels in an urban environment. As a consequence, there is a flourishing production of articles in the field of operational optimization of the combined use of trucks and drones for fulfilling customers requests. The aim is minimizing the total time required to se…
▽ More
The use of drones in logistics is gaining more and more interest, and drones are becoming a more viable and common way of distributing parcels in an urban environment. As a consequence, there is a flourishing production of articles in the field of operational optimization of the combined use of trucks and drones for fulfilling customers requests. The aim is minimizing the total time required to service all the customers, since this has obvious economical impacts. However in the literature there is not yet a widely recognized basic model, and there are not well assessed sets of instances and optimal solutions that can be considered as a benchmark to prove the effectiveness of new solution methods. The aim of this paper is to fill this gap. On one side we will clearly describe some of the most common components of the truck/drone routing problems and we will define nine basic problem settings, by combining these components. On the other side we will consider some of the instances used by many researchers and we will provide optimal solutions for all the problem settings previously identified. Instances and detailed solutions are then organized into benchmarks made publicly available as validation tools for future research methods.
△ Less
Submitted 28 July, 2021;
originally announced July 2021.
