A Topical Review of Quantum and Classical Machine Learning Approaches to Disaster Escape Routing Problems

The pathfinding problem in a graph has been solved using several classical algorithms, notably Dijkstra’s and A* algorithms. However, most classical algorithms are most effective on static graphs. They either cannot be adapted to dynamic graphs or become computationally expensive. The challenge of processing dynamic graphs, which demands significant computational resources, can be addressed using Classical and Quantum Machine Learning methods. For this review, we consider a dynamically changing graph representing a disaster-stricken city. Our problem, termed the Disaster Escape Routing problem, aims to find the optimal path within this dynamic graph. We review and analyze an existing hybrid quantum-classical machine learning model alongside classical machine learning models specifically for this problem. We also explore Variational Quantum Circuits and Encoding Methods. Our study suggests that hybrid quantum-classical machine learning, Graph Neural Networks, and Temporal Graph Networks offer high performance in terms of path prediction and accuracy in finding the optimal path. This review also identifies Kolmogorov Arnold Networks as a promising approach to solving escape routing problems. Additionally, an integrated approach combining strengths of all the models has been hypothesized to enhance emergency escape route planning.