Do We Trust Simulation Models?



Do We Trust Simulation Models?

Authors

CLEGG R G and CLUNE A, University of York, CASSIR C, University of Newcastle Upon Tyne, UK

Description

Transport planning involves the use of computer models as a tool for predicting the effects of real life changes to a traffic.network. However, relatively little work seems to be done on comparing the results of such models, and thus questioning their rel

Abstract

Transport planning involves the use of computer models as a tool for predicting the effects of real life changes to a traffic.network. However, relatively little work seems to be done on comparing the results of such models, and thus questioning their relative reliability.

This paper attempts to address this by comparing four different simulation models considering results on representations of the same road network. The models used are the well-known models SATURN and CONTRAM, the STEER model developed at the University of York and the Path Flow Estimator model developed at the University Of Newcastle. These models represent a spectrum of simulation modelling. STEER is a dynamic micro-simulation model whereas Path Flow Estimator is a static macro-simulation designed to run in real time.

Initial results obtained as part of the AIUTO project show that STEER, SATURN and PFE when simulating a real-life scheme, implemented on the York Network, produce results which, while broadly in agreement, differed considerably in the details. This paper will take this work further and answer the following questions:

* To what extent can the results of the four models studied be said to be in agreement?

* Are there any consistent levels of spatial aggregation where the models can be said to agree?

* Do different models respond differently to increased demand?

* When a transport planner asks questions of a model how much does the answer given depend on the model used?

* How well do the results of the modelling done agree with real life in each of the models?

* How clearly can the differences be attributed to, or explained by, the different principles underlying the models?

Publisher

Association for European Transport