Dynamic Strategies for Robust Rail Transit Operations against Disruptions by Malicious Attacks

Rail transit plays an important role in sustainable urban development. In Hong Kong, more than 10 million passenger trips are taken daily on the public transport system, and over 4 million, or 40%, of these trips are served by the local rail transit network operated by the MTR Corporation. In recent years, rail transit systems have been vulnerable under different disruptions. In particular, with the advances in information and communication technologies, malicious physical and cyberattacks now pose a real threat to our transport systems. We have seen various physical interruptions and attacks during the social unrest in 2019 that caused serve disruptions to MTR network in Hong Kong. These incidents could disrupt our transport systems and hence the smooth functioning of society. Following the Smart City Blueprint issued in 2020, several initiatives have been launched to improve the effectiveness and robustness of transit systems by using emerging technologies. Real time transit demand and operational data have now become increasingly available through automatic fare collection and vehicle positioning systems. Nevertheless, an effective computational framework is still lacking that can unlock the full potential of using these new data sources and technologies for improving the robustness of rail transit systems against malicious disruptions. This project will aim to develop a dynamic optimisation framework for modelling and optimising the operational decisions for rail transit systems with disruptions induced by malicious attacks. We first consider there is a system attacker aiming to impose disruptions that increase passengers’ travel times and operator’s costs. We then consider the system operator would seek the best corresponding actions to counteract the disruptions imposed by the attacker. The interaction and decisions of the attacker and system operator will be modelled through a dynamic attacker-defender game theoretic framework. Considering the computational complexity involved in the optimisation problem, the game theoretic framework is to be calculated and solved by using a reinforcement learning solution technique. The proposed reinforcement learning method reduces the computational complexity by making appropriate approximations during the calculation process of the original problem. The proposed models and algorithms will be implemented and tested using real-world scenarios generated from the Hong Kong MTR rail transit network. The effectiveness, robustness, and vulnerability of the underlying rail transit system under disruptions of different scales will be studied. This project will contribute to the development of advanced analytics and computational techniques for rail transit operations.