Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Restaurant Meal Delivery Problem with Ghost Kitchens

Gal Neria, Florentin D Hildebrandt, Michal Tzur, Marlin W Ulmer

TL;DR

The paper tackles the challenge of operating restaurant delivery under the ghost kitchen paradigm by formulating the Restaurant Meal Delivery Problem with Ghost Kitchens (RMD-GK) as a stochastic sequential decision process. It introduces an Anticipatory Integrated (AI) policy that jointly optimizes cook scheduling and vehicle dispatching using a Large Neighborhood Search (LNS) over a condensed decision representation, augmented with a Partial Decision Feasibility and Timing (PDFT) procedure and a Neural Network Value Function Approximation (VFA) with transfer learning. The results show that ghost kitchens, when operated with anticipatory, integrated optimization, yield significant improvements in delivery speed, bundling, and resource utilization compared to conventional setups and non-anticipatory baselines, while revealing a trade-off between fast delivery and freshness. The methodology offers a generalizable framework for complex, synchronized decision problems and provides managerial insights on resource balancing and collaboration between cooks and fleets. Overall, the work demonstrates that integrated, anticipatory scheduling and routing for ghost kitchens can meaningfully enhance service quality and efficiency in on-demand meal delivery systems.

Abstract

Restaurant meal delivery has been rapidly growing in the last few years. The main challenges in operating it are the temporally and spatially dispersed stochastic demand that arrives from customers all over town as well as the customers' expectation of timely and fresh delivery. To overcome these challenges a new business concept emerged, "Ghost kitchens". This concept proposes synchronized food preparation of several restaurants in a central complex, exploiting consolidation benefits. However, dynamically scheduling food preparation and delivery is challenging and we propose operational strategies for the effective operations of ghost kitchens. We model the problem as a sequential decision process. For the complex, combinatorial decision space of scheduling order preparations, consolidating orders to trips, and scheduling trip departures, we propose a large neighborhood search procedure based on partial decisions and driven by analytical properties. Within the large neighborhood search, decisions are evaluated via a value function approximation, enabling anticipatory and real-time decision making. We show the effectiveness of our method and demonstrate the value of ghost kitchens compared to conventional meal delivery systems. We show that both integrated optimization of cook scheduling and vehicle dispatching, as well as anticipation of future demand and decisions, are essential for successful operations. We further derive several managerial insights, amongst others, that companies should carefully consider the trade-off between fast delivery and fresh food.

The Restaurant Meal Delivery Problem with Ghost Kitchens

TL;DR

The paper tackles the challenge of operating restaurant delivery under the ghost kitchen paradigm by formulating the Restaurant Meal Delivery Problem with Ghost Kitchens (RMD-GK) as a stochastic sequential decision process. It introduces an Anticipatory Integrated (AI) policy that jointly optimizes cook scheduling and vehicle dispatching using a Large Neighborhood Search (LNS) over a condensed decision representation, augmented with a Partial Decision Feasibility and Timing (PDFT) procedure and a Neural Network Value Function Approximation (VFA) with transfer learning. The results show that ghost kitchens, when operated with anticipatory, integrated optimization, yield significant improvements in delivery speed, bundling, and resource utilization compared to conventional setups and non-anticipatory baselines, while revealing a trade-off between fast delivery and freshness. The methodology offers a generalizable framework for complex, synchronized decision problems and provides managerial insights on resource balancing and collaboration between cooks and fleets. Overall, the work demonstrates that integrated, anticipatory scheduling and routing for ghost kitchens can meaningfully enhance service quality and efficiency in on-demand meal delivery systems.

Abstract

Restaurant meal delivery has been rapidly growing in the last few years. The main challenges in operating it are the temporally and spatially dispersed stochastic demand that arrives from customers all over town as well as the customers' expectation of timely and fresh delivery. To overcome these challenges a new business concept emerged, "Ghost kitchens". This concept proposes synchronized food preparation of several restaurants in a central complex, exploiting consolidation benefits. However, dynamically scheduling food preparation and delivery is challenging and we propose operational strategies for the effective operations of ghost kitchens. We model the problem as a sequential decision process. For the complex, combinatorial decision space of scheduling order preparations, consolidating orders to trips, and scheduling trip departures, we propose a large neighborhood search procedure based on partial decisions and driven by analytical properties. Within the large neighborhood search, decisions are evaluated via a value function approximation, enabling anticipatory and real-time decision making. We show the effectiveness of our method and demonstrate the value of ghost kitchens compared to conventional meal delivery systems. We show that both integrated optimization of cook scheduling and vehicle dispatching, as well as anticipation of future demand and decisions, are essential for successful operations. We further derive several managerial insights, amongst others, that companies should carefully consider the trade-off between fast delivery and fresh food.
Paper Structure (51 sections, 1 theorem, 4 equations, 5 figures, 7 tables, 4 algorithms)

This paper contains 51 sections, 1 theorem, 4 equations, 5 figures, 7 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Problem: Meal Delivery Routing
  4. Model: Combined Scheduling and Routing
  5. Methodology: Search and Evaluation of Complex Decision Spaces
  6. Problem Statement
  7. Problem Description
  8. Example
  9. Sequential Decision Process
  10. Solution Method
  11. Motivation and Overview
  12. The Large Neighborhood Search
  13. A Condensed Representation of a Decision and a Partial Decision.
  14. Proof:
  15. Proof:
  16. ...and 36 more sections

Key Result

Theorem 4.1

Given a state $S_k$, an optimal decision can be found by searching over all (feasible) condensed decision representations rather than over all original decision representations.

Figures (5)

  • Figure 1: Overview of the steps of our AI-method in a decision state
  • Figure :
  • Figure A1: Termination percentage of the PDFT-algorithm with respect to the number of PDFT-iterations.
  • Figure A2: Percentage increase in delay for different number of LNS-iterations and AI policies compared to AI with 70 LNS-iterations.
  • Figure A3: Average workload utilization of vehicles and cooks by AI and FIFO over time, Large instances

Theorems & Definitions (17)

  • Definition 4.1: A condensed decision representation
  • Definition 4.2: Symmetric decisions
  • Claim 4.1: Condensed representation of symmetric decisions
  • Claim 4.2: Symmetric decision costs
  • Theorem 4.1
  • Definition 4.3: A partial decision
  • Claim A.3.1: Sequencing a set of similar orders
  • Claim A.3.2: Order departures by SPT
  • Claim A.4.2
  • Claim A.4.3
  • ...and 7 more