An Iterated Local Search Algorithm for the Pollution Traveling Salesman Problem
Motivated by recent works on the Pollution Routing Problem (PRP), introduced in Bektas and Laporte (Transp Res Part B: Methodol 45(8):1232–1250, 2011) , we study the Pollution Traveling Salesman Problem (PTSP). It is a generalization of the well-known Traveling Salesman Problem, which aims at finding a Hamiltonian tour that minimizes a function of fuel consumption (dependent on distance travelled, vehicle speed and load) and driver costs. We present a Mixed Integer Linear Programming (MILP) model for the PTSP, enhanced with sub-tour elimination constraints, and propose an Iterated Local Search (ILS) algorithm. It first builds a feasible tour, based on the solution of the Linear Programming (LP) relaxation of the MILP model, and then loops between three phases: perturbation, local search and acceptance criterion. The results obtained by the ILS on instances with up to 50 customers are compared with those found by a Cut-and-Branch algorithm based on the enhanced MILP model. The results show the effectiveness of the ILS algorithm, which can find the best solution for about \(99\%\) of the instances within short computing times.
KeywordsPollution Traveling Salesman Problem Iterated Local Search MILP model
This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550-17-1-0025. We also acknowledge project CONICYT FONDECYT 11150370.