Modelling negotiations in open railway access market for resource allocation

Abstract

An open railway access market usually consists of an infrastructure provider (IP) and a group of train service providers (TSPs). Through disintegration and distribution, the managerial responsibilities on railway resources are allocated to these stakeholders. To harmonise the interrelated resource allocation processes, negotiation among the stakeholders is an important and inevitable process to resolve the operational conflicts. This study aims to develop a software platform to enable such negotiations and investigate the behaviour of the stakeholders in negotiations. To address the distributed nature of the stakeholders and their behaviour during negotiation, a Multi-Agent System for Open Railway Access Market (MAS-ORAM) is established. MAS-ORAM is a virtual market where each stakeholder is represented by a software agent. With this system, the study focuses on modelling three major negotiations among the stakeholders (i.e. IP vs. TSP, IP vs. multiple TSPs, and TSP vs. TSP). Such representations of the open market and the subsequent study on the interactions between the stakeholders in railway management, particularly for open access markets, are the novelty of this research work. To facilitate rational decision-making by agents, not only has the study employed algorithms of standard optimisation techniques, but also of artificial intelligence approaches. The former includes Branch-and-Bound and Lemke’s Complementary Pivoting algorithms, while the latter involves the solving of a Prioritised Fuzzy Constraint Satisfaction problem. In addition, two policies on sequencing multiple negotiations by an IP and three negotiation strategies for TSPs have been devised. The performances of these algorithms and negotiation behaviours have been thoroughly examined through extensive simulation studies and statistical analysis. Simulation results have shown that software agents can be set up to represent railway stakeholders of different operation objectives. During negotiation, these agents exhibit rational behaviour in making concession. Results have also confirmed that the setup of MAS-ORAM is able to derive Pareto-optimal resource plans. The system, with proper enhancements and adaptation to a specific railway market, may thus be deployed by the open railway access market stakeholders as an analytical tool before the actual negotiations are conducted in practice. The research also demonstrates the feasibility of applying agent modelling in railway management.