Current students


CROCI DAVIDECycle: XXXVI

Section: Systems and Control
Advisor: JABALI OLA
Tutor: PIRODDI LUIGI

Major Research topic:
Heterogeneous service optimization in last mile logistics

Abstract:
The recent growth of electronic communication and web technologies has led to an irreversible decline in traditional delivery services, e.g., mail services. This same growth, however, has brought an exponential rise in the demand of home-based services like parcel delivery, grocery delivery or at-home health care services. Logistics companies operating traditional delivery services are adapting to those changes by gradually introducing the home-based services into their standard operations. The expansion of the range of offered services comes with new planning challenges. Thus, the topic of this thesis is the development of innovative algorithms to support logistics companies in transitioning towards home delivery services. The thesis is divided in three chapters.
At first, we formulate and solve a new districting problem to analyze the geographical distribution of the services of a logistics company. This problem takes as input the set of provided services over a finite time horizon and divides them in clusters based on their geographical position. Each cluster will be assigned to one or more deliverers, and the goal is to define clusters that minimize direct delivery costs while maintaining a fair workload distribution. We formulate this problem as a variant of the p-median problem and then propose a variable neighborhood search heuristic algorithm that can solve large problem instances by decomposing them in a series of sub problems.
In the second chapter we formulate and solve a new variant of the multi-period vehicle routing problem with due dates. We consider a set of heterogeneous home-based services, each having a different service level agreement. Introducing such services entails respecting services’ due dates spanning a number of days, and yields a complex operation that may require major changes to planned routes. We formulate this problem by establishing a priori routes to be operated on a daily basis, which are altered at minimal costs to account for the various due dates. After this formulation, we propose an adaptive large neighborhood search heuristic algorithm to solve larger problem instances.
The last part of this thesis is concerned with the extension of the previous routing algorithm to a stochastic and dynamic scenario. Most of the services demanded by the population nowadays need fast delivery. One example is the home delivery of food, where on average less than one hour passes between the order being placed and the actual delivery. The goal of this last chapter is the development of an algorithm that can optimize the introduction of these types of services in a pre-defined routing plan, even when demand is not known a-priori. By combining dynamic vehicle routing with revenue management, this algorithm determines whether to accept an incoming service request and at which price.