i-Deliver P&D Engine: A Decentralized Middleware for a Delivery-as-a-Service System

Abstract

We present the i-Deliver pickup-and-delivery (P&D) core algorithmic engine, a decentralized middleware which takes over the delivery logistics in an ecosystem of providers, consumers, couriers, and brokers who match orders for goods (e.g., food, parcels, shopping orders, etc.) between providers and consumers by appropriately (and automatically) allocating them to couriers in such a way that a global performance criterion (e.g., the total travel-time of couriers) is optimized. The middleware consists of two fundamental modules: (i) an order allocation module that assigns P&D orders to couriers in such a way that all spatiotemporal constraints for the orders, as well as constraints related to the couriers’ working conditions (e.g., working shifts) are respected; (ii) a courier guidance module that aids each courier by providing timely-service alerts and routing instructions, while servicing the orders assigned to them. We present the details for the P&D order allocation problem to be solved, via an appropriate MILP formulation. We then discuss both an offline algorithm whose goal is to solve this MILP in its entirety, and an online algorithm that handles the orders in real-time. The online solver (which can also be seen as a heuristic for solving the offline problem) is very suitable for servicing real-time orders, as demonstrated by the evidence provided by our thorough experimentation with real-world data.