A template‐based adaptive large neighborhood search for the consistent vehicle routing problem

Abstract

The importance of customer satisfaction was identified by many industries as a key factor of competitive advantage. So, for companies in the small package shipping industry, it can be reasonable to increase the service quality even at the expense of transportation cost to gain customer loyalty. These companies noticed that customer satisfaction can be increased by providing consistent service in the form of visiting customers with the same driver at approximately the same time of the day over a certain time period. Motivated by this real-world problem, the consistent vehicle routing problem (ConVRP) combines traditional vehicle routing constraints with the requirements for service consistency. This article presents a fast solution method called template-based adaptive large neighborhood search for the described problem. Compared to state-of-the-art heuristics, the developed algorithm is highly competitive on the available benchmark instances. Additionally, new test instances are provided. These seem to be more challenging due to the variation of different model parameters and consequently help to identify interesting effects. Finally, a relaxed variant of the original ConVRP is presented. In this variant, the departure times from the depot can be delayed to adjust the service times of the customers. Experiments show that allowing later departure times considerably improves the solution quality under tight consistency requirements.

The Generalized Consistent Vehicle Routing Problem

“…The template concept is also applied in Kovacs, Parragh, and Hartl (2014) and Tarantilis, Stavropoulou, and Repoussis (2012) to solve the ConVRP. In Tarantilis, Stavropoulou, and Repoussis, a tabu search algorithm is used to iteratively generate templates but also to improve the daily routes that are derived from the template routes.…”

“…State-of-the-art metaheuristics (Groër, Golden, and Wasil 2009;Tarantilis, Stavropoulou, and Repoussis 2012;Kovacs, Parragh, and Hartl 2014) rely on the template concept to generate good solutions. A template is a set of routes from which the daily routing plans are derived.…”

“…Flexible departure times are necessary because of the AM/PM time windows in combination with the constraints that prohibit waiting times (e.g., it is not possible to visit only PM customers if departure times are fixed at 0). Additionally, time consistency can be improved significantly by adjusting the vehicle departure times (Kovacs, Parragh, and Hartl 2014). The problem of adjusting the departure times of a given solution is modeled as follows.…”

“…The idea is to iteratively shift the departure time of the vehicle that visits the customer with the maximum arrival time difference. This approach is motivated by Kovacs, Parragh, and Hartl (2014). However, our algorithm is more sophisticated.…”

“…This extension is necessary because of the AM/PM time windows in combination with constraints that prohibit waiting times. Additionally, time consistency can be improved by adjusting the vehicle departure times (Kovacs, Parragh, and Hartl 2014).…”