MASSIEF: Modelling Disruption Management as a Multi-Agent System to Improve the Prediction of the Function Recovery Time

Abstract

When a disruption occurs on the rail track, the disruption management of ProRail is responsible for recovering the function of the failed infra object as safe and soon as possible, so the hindrance is minimalized. An important aspect of the disruption management is predicting the function recovery time (FRT). The four main parties of the disruption management faces challenges in estimating the FRT, due to decision making on invalidated information and the lack of information sources. The nature of this problem fits the characteristics of a multi-agent system (MAS) simulation. In the present study, I have built a MAS, which simulates the disruption management deterministically. I extended this baseline model with algorithmic modules and adjusted communication lines between the agents, which aimed to improved decision making on the predicted FRT. I have tested the extended MAS on five scenarios in which a switch was disrupted. The extended MAS predicted the FRT better than the original prediction in four of the five scenarios. The performance of the presented MAS is a proof of concept, showing that MAS modelling and extending the model, makes it able to generate a better prediction on the FRT.