Author:

Abstract:

This thesis considers three inventory problems, motivated by the transition toward greener logistics. We analyze how companies can make 'smart'' replenishment decisions, hence reducing the negative impact of their supply chains. First, we focus on perishable inventory management which, cursed by dimensionality, is notoriously difficult. This complexity forces retailers to resort to sub-optimal replenishment heuristics, leading to elevated costs and spoilage. We show how machine learning can be used to learn well-performing replenishment policies. By enriching the algorithm with advice from a known replenishment heuristic, we demonstrate that learning can be improved in terms of speed and performance. In the next chapter, we analyze how shippers can collaborate and shift part of their freight to a more sustainable transport mode, such as train, to reduce the dependency on (polluting) truck transport. We propose a heuristic for this collaborative, multi-modal replenishment problem and show how collaborative shipping can be a key to shift freight toward more sustainable transport modes in a cost-efficient way. In the last research chapter, we extend our analysis on collaborative shipping and consider a digital freight platform where excess transportation capacity is offered at a discount. We analyze the shipper's (complex) optimal response to these reduced rate offers, propose simple decision rules, and highlight when engaging with a platform is beneficial. Knowing the shipper's reaction to discount offers, we study the platform's pricing decision and find that a redistribution of the shipper's gains may be needed to ensure adequate platform usage.