Introduction

One could probably reach agreement that the traffic flow behavior of traffic simulation models should be well documented. Yet, in practice, this turns out to be somewhat difficult. Many traffic simulation models are under continuous development, and the traffic flow dynamics documented in a certain publication is often a ``snapshot'', valid at the time of writing, but no longer the true state of the model.

It thus makes sense to agree on a certain set of tests for traffic flow dynamics which should be run and documented together with ``real'' results. In this paper, we propose a (probably incomplete) suite of traffic flow measurements. Also, some of the results in this paper are arguably unrefined with respect to reality. Yet, as we stated above, we are continuously working on improvements, and this publication represents both a snapshot of where we currently stand and an argument for a standardized traffic flow test suite for simulation models. We hope that this publication will both open the way for a constructive dialogue on which standardized traffic flow tests should be run for traffic simulation models, and which of the features of our traffic simulation models may need improvement.

This paper starts with a general section on validation and calibration (Sec. 2), followed by a high-level description of the Transims microsimulation approach (Sec. 3). Sec. 4 is a fairly technical description of the actual implementation. Sec. 5 contains a description of the test cases that we ran for this paper and presents the simulation results. Sec. 6 contains an example of parameter sensitivity testing for the case of a yield sign, followed by a short section outlining differences in the logic when the simulation was used for the so-called Dallas case study (Sec. 7). The paper is concluded by a discussion section and a summary.