Exploring the Performance of Genetic Algorithm and Variable Neighborhood Search for Solving the Single Depot Multiple Set Orienteering Problem: A Comparative Study

TL;DR

This paper compares genetic algorithm and variable neighborhood search methods for solving the single depot multiple set orienteering problem, demonstrating that VNS outperforms GA in solution quality and efficiency, with ILP providing optimal solutions for small instances.

Contribution

It introduces and evaluates GA and VNS meta-heuristics for the sDmSOP, showing VNS's superior performance over GA and ILP for small problem sizes.

Findings

VNS outperforms GA in solution quality.

VNS requires less computation time than CPLEX.

ILP provides optimal solutions for small instances.

Abstract

This article discusses the single Depot multiple Set Orienteering Problem (sDmSOP), a recently suggested generalization of the Set Orienteering Problem (SOP). This problem aims to discover a path for each traveler over a subset of vertices, where each vertex is associated with only one cluster, and the total profit made from the clusters visited is maximized while still fitting within the available budget constraints. The profit can be collected only by visiting at least one cluster vertex. According to the SOP, each vertex cluster must have at least one of its visits counted towards the profit for that cluster. Like to the SOP, the sDmSOP restricts the number of clusters visited based on the budget for tour expenses. To address this problem, we employ the Genetic Algorithm (GA) and Variable Neighborhood Search (VNS) meta-heuristic. The optimal solution for small-sized problems is also suggested by solving the Integer Linear Programming (ILP) formulation using the General Algebraic Modeling System (GAMS) 37.1.0 with CPLEX for the sDmSOP. Promising computational results are presented that demonstrate the practicability of the proposed GA, VNS meta-heuristic, and ILP formulation by demonstrating substantial improvements to the solutions generated by VNS than GA while simultaneously needing much less time to compute than CPLEX.