Evaluating safety in a traffic simulation environment has always been a controversial research topic because vehicles in a simulation are programmed in such a way that they cannot potentially cause accidents. Alkis Papadoulis, PhD (Scientific Researcher at Aimsun) discusses how the i4Driving research project is aiming to address this issue by incorporating the most important human factors in existing human driver models; the hope is that the models will then serve as a reliable baseline for cooperative, connected and automated mobility.
According to the World Health Organisation, the number of road traffic deaths is “unacceptably high,” reaching 1.35 million every year worldwide. Road accGeidents are expected to become the fifth leading cause of deaths by 2030 (WHO, 2015). Because of this grim projection, academia, original equipment manufacturers, the automotive industry, and highway infrastructure providers have focused their efforts on developing both infrastructure-based and vehicle-based technologies that aim to improve road safety.
Cooperative, connected and automated mobility (CCAM) is one of the most recent examples of such technology that has advanced significantly over the past few years, and an increasing number of real-world testsare taking place world-wide to test its readiness and effectiveness. These tests have provided useful insights with regards to CCAM performance; however, they can come at a heavy price. Early-stage, real-world CCAM tests have caused accidents, which in some cases have even cost the lives of drivers or pedestrians.
Traffic simulation could be a good alternative virtual environment to test CCAM technologies. However, safety evaluation in traffic simulation has always been a controversial topic among transport researchers and has received criticism due to the overarching idealized human driver models that are expected to interact with Connected and Automated Vehicles (Tarko, 2005). When looking at existing simulation environments, one faces an important paradox: How can road safety be evaluated when in this virtual environment certain safety critical events that can potentially occur – rarely but they do occur – are not able to be reproduced? To be more specific, vehicles in current traffic simulation environments cannot crash and important human factors related to human reaction times, distraction and perception errors are not captured.
Just to be clear, seeing vehicles crash in every single simulation is not what I am suggesting, they just need to have the potential to do so. In any case, crashes, or accidents as they are more commonly referred to in literature, are the tip of the iceberg, or the tip of the pyramid (see below). More importantly, vehicles need to be able to reproduce realistic numbers of safety critical behaviors which may result in events way more common than crashes, such as near-misses (sometimes also known in road safety terms as ‘conflicts’ in Hyden’s safety pyramid shown below). Only then can simulation environments provide a reliable virtual testbed for CCAM technology.