[Paper Review] A Comprehensive Survey on the Ambulance Routing and Location Problems

Joseph P. Tassone, Salimur Choudhury|arXiv (Cornell University)|Jan 10, 2020

Facility Location and Emergency Management98 references4 citations

TL;DR

This paper presents a comprehensive survey of ambulance routing and location problems (ARP and ALP), framing them as NP-hard variants of the vehicle routing problem (VRP) and maximum coverage problem (MCP). It reviews existing methodologies, including metaheuristics and simulation, and identifies gaps in hybrid algorithms and machine learning applications, particularly reinforcement learning, for optimizing emergency medical services.

ABSTRACT

In this research, an extensive literature review was performed on the recent developments of the ambulance routing problem (ARP) and ambulance location problem (ALP). Both are respective modifications of the vehicle routing problem (VRP) and maximum covering problem (MCP), with modifications to objective functions and constraints. Although alike, a key distinction is emergency service systems (EMS) are considered critical and the optimization of these has become all the more important as a result. Similar to their parent problems, these are NP-hard and must resort to approximations if the space size is too large. Much of the current work has simply been on modifying existing systems through simulation to achieve a more acceptable result. There has been attempts towards using meta-heuristics, though practical experimentation is lacking when compared to VRP or MCP. The contributions of this work are a comprehensive survey of current methodologies, summarized models, and suggested future improvements.

Motivation & Objective

• To provide a systematic review of recent advancements in ambulance routing and location problems (ARP and ALP), which are critical for emergency medical service (EMS) efficiency.
• To analyze the adaptation of VRP and MCP methodologies to EMS-specific constraints, such as response time urgency and dynamic emergency demands.
• To identify research gaps in current approaches, particularly the underuse of hybrid metaheuristics and machine learning techniques like reinforcement learning.
• To evaluate the limitations of simulation-based methods and advocate for their use as preliminary tools rather than primary optimization strategies.
• To guide future research by proposing improvements in algorithmic integration, especially in combining metaheuristics and emerging AI techniques for real-time EMS optimization.

Proposed method

• Conducting a comprehensive literature review of ARP and ALP from 2000 to 2020, focusing on optimization techniques and modeling approaches.
• Classifying existing methods into categories: exact mathematical programming (e.g., integer linear programming), metaheuristics (e.g., GA, PSO, ACO, Tabu Search), and simulation-based approaches.
• Mapping ARP and ALP to their parent problems—VRP and MCP—by identifying shared structures, such as graph-based formulations G=(V,E), and adapting their objective functions and constraints.
• Evaluating the performance of metaheuristics, particularly noting that discrete encoding methods outperform continuous ones in ARP and ALP applications.
• Exploring the potential of reinforcement learning and deep reinforcement learning for dynamic EMS optimization, emphasizing state-action formulation and reward maximization.
• Assessing the feasibility of hybrid algorithms that combine multiple metaheuristics, highlighting challenges in compatibility and implementation complexity.

Experimental results

Research questions

• RQ1 How do ambulance routing and location problems generalize the vehicle routing problem (VRP) and maximum coverage problem (MCP), and what are the key modifications in objectives and constraints?
• RQ2 What are the dominant optimization techniques used in ARP and ALP, and how do they compare in terms of scalability and solution quality?
• RQ3 Why is simulation widely used in current research, and what are its limitations as a standalone optimization method for EMS systems?
• RQ4 What are the major gaps in current research, particularly regarding the integration of machine learning and hybrid metaheuristic approaches?
• RQ5 How can reinforcement learning and deep learning be effectively applied to dynamic ambulance routing and location decisions, and what are the main implementation challenges?

Key findings

• ARP and ALP are NP-hard problems, necessitating the use of metaheuristics or approximate methods for large-scale real-world applications.
• Simulation is frequently used but is not a substitute for true optimization, as it primarily validates systems rather than discovering optimal solutions.
• Metaheuristics such as Genetic Algorithms, Particle Swarm Optimization, and Ant Colony Optimization are widely applied, with discrete encoding showing superior performance in routing and location models.
• There is a notable lack of hybrid algorithms combining multiple metaheuristics, despite their proven success in related problems like VRP.
• Machine learning applications, particularly reinforcement learning, remain underexplored but offer significant potential due to their ability to learn from state-action-reward dynamics without requiring labeled training data.
• Deep reinforcement learning, combining convolutional neural networks with Q-learning, shows promise for solving complex, dynamic EMS optimization problems, though training time and proper state-action encoding remain major challenges.