Last-mile Analytics

Reinforcement Learning Approaches for the Orienteering Problem with Stochastic and Dynamic Release Dates

Speaker: Claudia Archetti

Abstract: We study Orienteering Problem with Stochastic and Dynamic Release Dates (DOP-rd) where a single and uncapacitated vehicle is serving customer requests with stochastic and dynamic release dates, representing the time at which parcels are available for distribution. The objective is to maximize the number of requests served within the deadline. We model the problem as a Markov decision process and present two approximation approaches for its solutions, where the distribution strategy is learned from Monte-Carlo simulation.

Online Approaches for the Drone-Supported Distribution of Relief in Disaster Areas

Speaker: Marie Schmidt

Abstract: We study the problem of providing emergency supplies in disaster areas by a tandem of truck and drone. In the basic problem version, we assume that we are given a number of potential demand points. Upon visiting a potential demand point by either vehicle, we learn whether there is indeed demand for relief goods in the respective node. In that case a visit by truck is necessary to deliver the required goods. We propose a routing strategy for truck and drone and are able to prove that it has a competitive ratio of 2 if the truck is faster than the drone and 1+alpha if the drone is faster, where alpha denotes the ratio of truck and drone speed. Furthermore, for some values of alpha, we are able to show that the proposed strategy is optimal in the sense that no other strategy can achieve a better competitive ratio.

Assessing Economic Performance of E-grocery Delivery

Speaker: Moritz Fleischmann

Abstract: Industry experts and scholars commonly consider grocery home delivery (e-grocery) in developed economies as unprofitable. Nevertheless, both the number of companies offering e-grocery services and the share of online grocery sales continue to grow. Common justifications for pursuing currently unprofitable business activities rely  on growth arguments. In this talk, we investigate the impact of growth on the profitability of e-grocery services. To this end, we develop an analytic model that captures the key profit drivers and calibrate it using publicly available real-life data. At the core of our model is the economic structure of the delivery operations which we model using continuous approximation.