[Paper Review] A supervised hybrid quantum machine learning solution to the emergency escape routing problem

Nathan Haboury, Mo Kordzanganeh+7 more|arXiv (Cornell University)|Jul 28, 2023

Traffic Prediction and Management TechniquesEngineering10 citations

TL;DR

The paper presents a hybrid quantum-classical supervised learning approach to mimic node-wise Dijkstra’s routing on evolving earthquake-affected city graphs, achieving higher accuracy than a purely classical model and feasible quantum hardware execution.

ABSTRACT

Managing the response to natural disasters effectively can considerably mitigate their devastating impact. This work explores the potential of using supervised hybrid quantum machine learning to optimize emergency evacuation plans for cars during natural disasters. The study focuses on earthquake emergencies and models the problem as a dynamic computational graph where an earthquake damages an area of a city. The residents seek to evacuate the city by reaching the exit points where traffic congestion occurs. The situation is modeled as a shortest-path problem on an uncertain and dynamically evolving map. We propose a novel hybrid supervised learning approach and test it on hypothetical situations on a concrete city graph. This approach uses a novel quantum feature-wise linear modulation (FiLM) neural network parallel to a classical FiLM network to imitate Dijkstra's node-wise shortest path algorithm on a deterministic dynamic graph. Adding the quantum neural network in parallel increases the overall model's expressivity by splitting the dataset's harmonic and non-harmonic features between the quantum and classical components. The hybrid supervised learning agent is trained on a dataset of Dijkstra's shortest paths and can successfully learn the navigation task. The hybrid quantum network improves over the purely classical supervised learning approach by 7% in accuracy. We show that the quantum part has a significant contribution of 45.(3)% to the prediction and that the network could be executed on an ion-based quantum computer. The results demonstrate the potential of supervised hybrid quantum machine learning in improving emergency evacuation planning during natural disasters.

Motivation & Objective

• Motivate optimized emergency evacuation routing under dynamic earthquake-induced traffic changes.
• Model the routing problem as a dynamic graph and reformulate as a node-wise shortest-path task guided by Dijkstra’s algorithm.
• Develop a hybrid FiLM-based architecture combining classical and quantum components to imitate shortest-path decisions.
• Evaluate whether the quantum component meaningfully contributes to inference and assess hardware feasibility on ion-based QPUs.

Proposed method

• Model the evacuation problem as a dynamic weighted graph with evolving earthquake and traffic effects.
• Train a hybrid quantum-classical supervised model to imitate node-wise Dijkstra’s decisions using labeled data from Dijkstra’s algorithm.
• Use a FiLM neural network to condition on earthquake coordinates and a parallel quantum neural network (PHN) to process state features.
• Implement a 7-qubit variational quantum circuit with data reuploading for earthquake inputs and a five-qubit main circuit for graph features.
• Combine quantum and classical outputs via a final fully connected layer to predict the next node.
• Evaluate with metrics of arrival rate and accuracy against node-wise Dijkstra’s results.

Figure 1: The mathematical abstraction of the emergency escape routing problem during an earthquake. The map of Furubira was investigated by converting it to a computational graph. Each node represents a crossroad where a decision on the direction is required. The three exit points on the map are ex

Experimental results

Research questions

• RQ1 Can a hybrid quantum-classical model imitate node-wise Dijkstra’s routing on dynamically changing city graphs during earthquakes?
• RQ2 Does the quantum component meaningfully contribute to prediction accuracy and path quality in emergency routing tasks?
• RQ3 Is the hybrid model performant enough to run on contemporary ion-based quantum hardware for short-path decisions?
• RQ4 What input features and data engineering choices maximize learning of shortest-path decisions under evolving conditions?

Key findings

• HQNN achieves an average accuracy of 94% compared to 87% for the classical neural network.
• HQNN reaches a higher arrival rate and more paths that are equal to or faster than Dijkstra’s results.
• The quantum part contributes meaningfully to inference with a relative quantum contribution of 0.45(3) in the final layer.
• The model can be executed on an ion-based quantum computer for short paths, with qualitative agreement to simulator results.
• Hybrid model shows robustness to dynamic graph changes and can outperform classical in certain evolving environments.

Figure 2: The supervised learning approach to the emergency escape routing problem. The hybrid quantum neural network (HQNN) trains to imitate the node-wise Dijkstra’s algorithm.