The paper presents recent examples of the use of two methods in assessing the safety effects of the so-called intelligent vehicle safety.uThese methods were applied to three in-vehicle ITS systems. Estimates produced by the different approaches differ considerably and would lead to considerable variation in e.g. selecting the priority systems to be promoted in national and international safety programmes. The paper discusses the differences in the estimates as produced by the different approaches and attempts to identify the factors behind these differences. The paper also discusses the need to develop the ex-ante assessment of the safety effectiveness of systems. There are many differences between the estimates of the safety effects of a specific ITS application. A part of these differences is naturally linked to the circumstances and assessment scenarios such as the country, road class, weather conditions, traffic volumes etc. But in a number of instances large differences are caused by the fact that some assessments do not consider all possible relevant safety mechanisms. We can investigate these omissions also from the perspective of the three safety dimensions. Concerning consequences and systems only affecting the consequences of the accident, either during the crash or after it, it seems that the main effects are estimated in a very similar manner when applying either the accident causation or behavioural effect approach. When systems are affecting accident or crash risk, there seem be differences based on the approach. The primary effects related to crash risk are usually always covered with both approaches, but in many cases all of the other effects are not covered. The risk for not covering some effects may be greater with the accident causation approach, and especially with behavioural adaptation effects. The same holds also for systems and applications affecting exposure. Even with the behavioural effect approach, there is a risk of not covering the effects on exposure. It should also be pointed out that both approaches are very sensitive to assumptions made in the assessment process. Assumptions have to be made in all cases as all information required is very rarely available. When using the behavioural effect approach, researchers tend to have some information of the application's effect on driving and travelling behaviour on the basis of simulator, modelling or field studies. This can then be transformed to the overall accident/injury/fatality effectiveness of the application. When using the accident causation approach, the effectiveness assumptions used are in many cases just guesses, although often based on earlier accident statistic based effectiveness estimates of similar types of applications. Here, the time horizon approach as proposed by Abele et al. (2004) could be useful in adding to the transparency of the accident causation approach. However, as such both methods are as good and usable, but much care needs to be taken in applying the methods. It would seem extremely useful, if the researchers applying the accident causation approach would utilise the nine-point behavioural checklist of Draskoczy, Carsten and Kulmala (1998). Similarly, the researchers utilising the behavioural approach would benefit from studying the accident data with the same care as those applying the accident causation approach. All behavioural effects need to be transformed into accident/injury/fatality effects in any case in the end to assess the effects on safety. Hence, both approaches should be seen as complementing approaches rather than competing ones. For the covering abstract see ITRD E138952. This paper is available from http://www.ictct.org/workshops/06-Minsk/Kulmala.pdf.
Samenvatting