Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cloud Kitchen: Using Planning-based Composite AI to Optimize Food Delivery Processes

Slavomír Švancár, Lukáš Chrpa, Filip Dvořák, Tomáš Balyo

TL;DR

The paper addresses optimizing restaurant delivery with their own fleets by modeling the problem as VRPTW and applying a planning-based composite AI approach. Cloud Kitchen combines a Technology-Specific Bridge, a Decision-Making component solving VRPTW tasks within UPF, a PDDL representation for explainability, and a SimPy-based simulator to evaluate decisions on historical data. Empirical evaluation shows substantial reductions in delivery delays (roughly one third) and in deliveries delayed by more than 10 minutes (about 40%), with modest increases in driving time and distance. The work demonstrates the practical value of compositional AI for logistics planning and provides a deployable blueprint for explainable routing in restaurant delivery contexts.

Abstract

The global food delivery market provides many opportunities for AI-based services that can improve the efficiency of feeding the world. This paper presents the Cloud Kitchen platform as a decision-making tool for restaurants with food delivery and a simulator to evaluate the impact of the decisions. The platform contains a Technology-Specific Bridge (TSB) that provides an interface for communicating with restaurants or the simulator. TSB uses a planning domain model to represent decisions embedded in the Unified Planning Framework (UPF). Decision-making, which concerns allocating customers' orders to vehicles and deciding in which order the customers will be served (for each vehicle), is done via a Vehicle Routing Problem with Time Windows (VRPTW), an efficient tool for this problem. We show that decisions made by our platform can improve customer satisfaction by reducing the number of delayed deliveries using a real-world historical dataset.

Cloud Kitchen: Using Planning-based Composite AI to Optimize Food Delivery Processes

TL;DR

The paper addresses optimizing restaurant delivery with their own fleets by modeling the problem as VRPTW and applying a planning-based composite AI approach. Cloud Kitchen combines a Technology-Specific Bridge, a Decision-Making component solving VRPTW tasks within UPF, a PDDL representation for explainability, and a SimPy-based simulator to evaluate decisions on historical data. Empirical evaluation shows substantial reductions in delivery delays (roughly one third) and in deliveries delayed by more than 10 minutes (about 40%), with modest increases in driving time and distance. The work demonstrates the practical value of compositional AI for logistics planning and provides a deployable blueprint for explainable routing in restaurant delivery contexts.

Abstract

The global food delivery market provides many opportunities for AI-based services that can improve the efficiency of feeding the world. This paper presents the Cloud Kitchen platform as a decision-making tool for restaurants with food delivery and a simulator to evaluate the impact of the decisions. The platform contains a Technology-Specific Bridge (TSB) that provides an interface for communicating with restaurants or the simulator. TSB uses a planning domain model to represent decisions embedded in the Unified Planning Framework (UPF). Decision-making, which concerns allocating customers' orders to vehicles and deciding in which order the customers will be served (for each vehicle), is done via a Vehicle Routing Problem with Time Windows (VRPTW), an efficient tool for this problem. We show that decisions made by our platform can improve customer satisfaction by reducing the number of delayed deliveries using a real-world historical dataset.
Paper Structure (7 sections, 3 figures, 1 table, 1 algorithm)

This paper contains 7 sections, 3 figures, 1 table, 1 algorithm.

Table of Contents

  1. Vehicle Routing Problem with Time Windows
  2. Cloud Kitchen Platform
  3. Decision-Making Component
  4. Technology-Specific Bridge
  5. Simulator
  6. Empirical Evaluation
  7. Conclusion

Figures (3)

  • Figure 1: Architecture of the Cloud Kitchen platform
  • Figure 2: A screenshot of the recommender. On the left-hand side, the recommender shows the "grouped" deliveries (with not-yet-ready orders at the bottom). On the right-hand side, the recommender depicts the delivery routes in maps.
  • Figure 3: Comparison of the numbers of orders per day delivered later than 10 minutes ($y$-axis) in historical data and planned by the Cloud Kitchen platform. Particular days are on the $x$-axis.