Towards a conflict prevention strategy applicable for real-time railway traffic management

摘要

In real-time, dispatchers have the challenging task to perform railway traffic management. However, it is almost impossible to anticipate the impact of their actions on the entire network. In order to support dispatchers in making decisions, a Traffic Management System (TMS) is required capable of predicting train movements, detecting conflicts and deciding on how to prevent these conflicts. This paper presents a Conflict Prevention Strategy (CPS) that can easily be integrated into an industrial TMS. We base on previous research on CPS which included a retiming/reordering heuristic and a local rerouting optimization, and a 'dynamic-impact-zone' heuristic to determine impact of conflicts. This paper proposes a series of algorithmic improvements to design a new strategy, called iCPS, usable in practice and able to solve very large and complex networks, very quickly, and with good performance. Moreover, the degree of realism of the instances can be increased, including multiple business rules related to passenger delays, and canceling trains in case of very large delays. This iCPS is tested on a large study area, and validated on an even larger network. The proposed iCPS outperforms the FCFS strategy by, on average, 51% in train delays and 22% in passenger delays while delivering solutions in 2.4 s on average and in 34.1 s at maximum. It also outperforms the previous CPS by up to 11% in solution quality.