The lateral direction of traffic, introduced into modelling during the last decade, is a matter of growing interest. If properly developed, 2D traffic models may allow more accurate reproduction of traffic, thus enhancing the effectiveness and application field of traffic simulation, as well as the understanding of driver behaviour. A more complete knowledge of human driving, in turn, is an important requirement for human-friendly autonomous vehicles sharing streets with human drivers. This paper addresses a tendency of drivers we call collateral anomaly (CA), by which they tend to choose different lateral positions depending on whether they are accompanied side by side by another vehicle or not. Although intuitive, it has never been measured before. In this work we characterise it by statistical means starting from three real trajectory datasets belonging to the Next-Generation Simulation project. We found evidence that the lateral position of vehicles follows a slightly different distribution when they have another vehicle vis-à-vis, deviating from the expected distribution by more than 50 mm on average. The effect is correlated to but cannot be explained alone by lane-changes, and seems to increase with speed. This kind of measure is a necessary step towards analysing 2D traffic model performance. Understanding CA might help understand other lateral position-related phenomena, like the relaxation phenomenon, or the influence of lane and vehicle width in traffic.
Nota bibliográficaPublisher Copyright:
© 2015 Hong Kong Society for Transportation Studies Limited.