Logistics of Surplus Food Rescue and Distribution

Abstract

Not-for-profit food rescue organisations play a vital role in alleviating hunger in many developing and developed countries. They rescue surplus food from the business sector and re-distribute it to welfare agencies supporting different forms of food relief. A food rescue organisation’s potential to balance both collection and delivery requirements is challenging, particularly when the network is diverse, resources are limited, and characteristics of the rescued food items influence the collection and delivery process. This thesis presents various statistical and mathematical models to facilitate the sustainable, efficient, fair and cost-effective flow of rescued surplus food products from point of donation to point of delivery. Various statistical modelling techniques including multiple linear regression, structural equation modelling and neural networks are used to explore the patterns and dynamics of food donation and the distribution process. Results suggest that structural equation modelling and neural networks provide better estimations for each food type when compared to conventional multiple linear regression. In addition, this thesis aims at exploring the potential of vehicle routing, scheduling and resource allocation models to address the concerns of food rescue and delivery operations. Specifically, in incorporating the perishability of products, level of fairness in the distribution of limited rescued food, uncertainty in supply and the number of food providers, etc., in addition to the traditional cost-effective operation, which are highly relevant to the operational policies of food relief logistics. The models are formulated by extending the unpaired pickup and delivery vehicle routing problem, variants of the classical vehicle routing problem. Application of these models in real case scenarios suggests a reduction in the overall operational cost along with a reduction in waste and improvement in fairness. The core contribution of the research is the development of novel mixed integer linear programming models for the scheduling, routing and allocation problems and computationally efficient Tabu Search based heuristic solution approaches to solve large scale instances. Furthermore, the research demonstrates the impact of these models in real case scenarios and implements them using significantly fewer resources than what are employed in practice.