Variable Neighborhood Search for the Multi-Depot Multiple Set Orienteering Problem

TL;DR

This paper proposes a Variable Neighborhood Search approach for a new multi-depot, multi-set orienteering problem, demonstrating improved solution quality and efficiency over exact methods across various instance sizes.

Contribution

It introduces a novel formulation for the multi-Depot multiple Set Orienteering Problem and applies a VNS meta-heuristic, showing its effectiveness on benchmark instances.

Findings

VNS outperforms exact solvers in solution quality and computational time.

Increasing the number of travelers significantly improves profit collection.

The proposed method is robust across small, medium, and large instances.

Abstract

This paper introduces a variant of the Set Orienteering Problem (SOP), the multi-Depot multiple Set Orienteering Problem (mDmSOP). It generalizes the SOP by grouping nodes into mutually exclusive sets (clusters) with associated profits. Profit can be earned if any node within the set is visited. Multiple travelers, denoted by $t$ $(> 1)$, are employed, with each traveler linked to a specific depot. The primary objective of the problem is to maximize profit collection from the sets within a predefined budget. A novel formulation is introduced for the mDmSOP. The paper utilizes the Variable Neighborhood Search (VNS) meta-heuristic to solve the mDmSOP on small, medium, and large instances from the Generalized Traveling Salesman Problem (GTSP) benchmark. The results demonstrate the VNS's superiority in robustness and solution quality, as it requires less computational time than solving the mathematical formulation with GAMS 37.1.0 and CPLEX. Additionally, increasing the number of travelers leads to significant improvements in profits.