System Simulation and Modelling

Where real-world testing proves either too expensive or too risky, the use of system simulation and modelling can provide a useful alternative. For example, methods such as systems dynamics (Sterman 2000) and agent-based modelling have previously been used to simulate the impacts of changes within transportation systems (e.g. Goh and Love 2012, McClure et al. 2015, Shepherd 2014).

Agent-based modelling is a computer modelling approach in which rules are applied to direct the behaviour of individual agents who then interact, enabling the identification of emergent properties arising from the interactions.

An agent-based model typically comprises the following three elements (Macal and North 2010):

  • 1. A set of agents, each with attributes and behaviours
  • 2. A set of agent relationships and methods of interaction
  • 3. The agents’ environment

This enables the analyst to model both, interactions between agents and interactions between agents and their environment. The modelling therefore provides a bottom- up analysis of system functioning and enables analysts to identify how changes to the rules governing interactions lead to emergent behaviours and affect the system as a whole. Agent-based modelling techniques have previously been used to model transportation systems (Sarvi and Kuwahara 2007, Thompson et al. 2015).

Systems dynamics, another computation modelling technique, is generally used to understand the implications of wider system change, such as policy change. Models are usually represented in diagrammatical form such as stock and flow diagrams with the relationships between factors in the model underpinned by differential equations (Hettinger et al. 2015). Both agent-based modelling and systems dynamics have limitations, and thus adaptations or hybrid approaches have been proposed to address these issues (e.g. Hettinger et al. 2015, Thompson et al. 2015). Importantly, it is proposed that these types of modelling techniques could be directly informed by HTA and CWA outputs. For example, the values and priority measures in an abstraction hierarchy could become variables in systems dynamics models, whereas the strategies identified in the Strategies Analysis phase can help to identify the behaviours of agents well as the rules or conditions under which those behaviours would be more likely to occur.

Of course, the validity of any modelling approaches must be addressed, and the outcomes will rely on the quality of the inputs and the assumptions made by analysts. Yet, these approaches provide a promising way to evaluate and compare options, and to identify potential issues not previously recognised.

A strength of computational modelling approaches is that they analyse system functioning over time. The approaches taken in the current research programme described rail level crossing systems at a point in time, but given that road and rail networks are dynamic open systems, these models may not remain accurate for long. The increasing uptake of new technologies in transportation systems, including higher forms of automation in both road vehicles and trains, raises questions about how system functioning will change in future. Furthermore, it is unknown how the recommended rail level crossing design changes might interact with a more highly automated transport system. Modelling techniques may assist to test assumptions about future systems and how interactions might change over time.

< Prev   CONTENTS   Source   Next >