Solving the One-Commodity Pickup and Delivery Problem Using an Adaptive Hybrid VNS/SA Approach

Abstract

In the One-Commodity Pickup and Delivery Problem (1-PDP), a single commodity type is collected from a set of pickup customers to be delivered to a set of delivery customers, and the origins and destinations of the goods are not paired. We introduce a new adaptive hybrid VNS/SA (Variable Neighborhood Search/Simulated Annealing) approach for solving the 1-PDP. We perform sequences of VNS runs, where neighborhood sizes, within which the search is performed at each run, are adaptable. Experimental results on a large number of benchmark instances indicate that the algorithm outperforms previous heuristics in 90% of the large size test cases. Nevertheless, this comes at the expense of an increased processing time.

Chapter 6: Pickup-and-Delivery Problems for Goods Transportation

“…Their computing times (reported only for the large instances) were comparable to those of Hernández-Pérez, Rodríguez-Martín, and Salazar-González [58]. Hosny and Mumford [64] obtained further improvements by using a Variable Neighborhood Search (VNS) algorithm, but at the expense of very high computing times (up to 42 hours on the large instances). To our knowledge, the best computational results have been obtained by Mladenović et al [92].…”

A one-commodity pickup-and-delivery traveling salesman problem solved by a two-stage method: A sensor relocation application

“…In it, 2-opt is used to accelerate the convergence of the GA after a pheromone-based crossover operator constructs offspring by utilizing pheromone trails and some local information including edge lengths and demands of customers to construct offspring. Hosny et al [28] introduced a new adaptive hybrid variable neighborhood search with a simulated annealing approach. The neighborhood size is adaptive with the current stage of the search in each run.…”

Adaptive search techniques for problems in vehicle routing, part II: A numerical comparison

“…For adapting the maximum neighborhood size in (A.1), we follow Hosny et al [9]. The other cases (A.2) -(A.6) are solved and compared with two different adaptive mechanisms: The first adaptive mechanism of the VNS is performed with a scoring system.…”

“…A simple adaptive strategy is presented by Hosny et al [9]. Besides an adaptable stopping condition controlled by the number of non-improving iterations, the maximum neighborhood size κ is not fixed, but it depends on the stage of the current VNS run considering multiple VNS runs.…”