Quantification of safe driving

Quantification of safe driving

Souman, J.L.
Adjenughwure, K.
van Dam, E.
van Weperen, M.
Tejada, A.

2021

This report summarizes the findings of the QuaSaR project that TNO conducted for the Netherlands Ministry of Infrastructure and Water Management (I&W). The Ministry has asked for this information to determine the feasibility of defining a method to quantitatively assess driving safety that could be applied to driving with automated vehicles. Assessing driving safety calls for the quantification of driving performance in relationship to safety. While traffic safety can be evaluated by measuring the number of fatalities, injuries and crashes in traffic, this is only possible after these crashes have happened. The challenge in measuring driving safety is to relate driving behaviour as measured to the likelihood of crashes occurring. This not only concerns the driving behaviour of one single driver and vehicle, but also the interactions of this driver with other road users. Consequently, not only driving behaviour but also traffic interactions need to be quantified and related to safety. In this report we provide a summary of the state of the art in qualitative and quantitative methods to reason about the safety of driving. To this end, we provide an overview of the following methods: Surrogate Safety Metrics, DOCTOR method, measures of driving impairment, data-driven methods, driving examination and driver tests (for human drivers). This information is complemented with insights obtained from interviewing three experts on driving examination and/or driving instruction. All interviewees had many years of experience in driver training and assessment. With these interviews, we hoped to gain insights in the assessment procedures for human drivers and in the experts’ views on the assessment of automated vehicles. The methods and insights explained above were analysed to determine whether they could be used as part of a safe driving assessment method for automated vehicles. To this end we developed a first set of requirements for such a method and discussed the relative benefits of all the techniques we reviewed. Importantly, we focused on methods that can be applied to the evaluation of driving safety of automated vehicles, without reference to properties of the driver (such as attention, anticipation or fatigue). Our main conclusion is that there is no consensus in the literature on how to assess safe driving. There are also no clear sets of requirements for such methods for either human-driven or automated vehicles. We recommend investigating these requirements further, especially for automated vehicles. We also recommend establishing the relationship between current quantitative measures of driving performance and traffic interactions with existing driving safety assessment criteria by human experts. Finally, effort should also be placed into developing an approach to an assessment that takes into account the wide variety of traffic situations and environmental conditions.

Automated vehicles
Safety
Surrogate safety metrics
Doctor method
Driving impairment
Data-driven methods
Driving examination
Driver tests
Assessment criteria

http://resolver.tudelft.nl/uuid:df4404be-252b-4187-832e-28dfd1609c4b

962160

TNO 2021 R12632 0.1

TNO, Helmond

report