A model library of driving behaviours
Before autonomous vehicles can hit the road, they first need to prove that they are as safe as human drivers. “The safe and efficient coexistence of automated and human-driven vehicles poses a profound, interdisciplinary challenge to transportation science,” says Maria Rodrigues, senior project manager for transport and mobility at Panteia. To meet this challenge, transportation science relies heavily on virtual simulations to run tests. But to work, these simulations need to be built on credible models of human driving behaviour. This is where the EU-funded i4Driving project comes in. “Our goal is to develop a new model library of human driving behaviour that will serve as a human baseline for road safety, against which automated driving systems (ADSs) and functions can be compared and evaluated,” explains Vincenzo Punzo, i4Driving scientific coordinator and professor at the University of Naples Federico II in Italy.
Assessing critical situations
For the i4Driving project, a new library doesn’t necessarily mean new models. Instead, it means developing a combination of new and existing models that can be used to assess ADSs in both scenario- and traffic-based safety simulations. According to Punzo, such combinations will ensure that virtual simulations accurately reflect the complexity of driving behaviour in realistic road traffic systems. “On the one hand, by incorporating the heterogeneity of drivers and driver behaviours, our models can mimic both the non-critical and safety-critical situations found in daily traffic,” he adds. “On the other hand, ‘proper’ system complexity is needed to make a robust and meaningful analysis of road safety.” Although the project is still ongoing, several important results have already been achieved. Groundbreaking experiments based on the models have been carried out on four driving simulations. The experimental design combines driver and environmental characteristics – such as age, gender, driving experience or lighting conditions. They also use different operational design domains – urban and extra-urban – in different locations across Europe. “Unlike traditional driving simulator experiments, which investigate specific scenarios, i4Driving’s experiments expose a driver to complex road and traffic sequences similar to those encountered in naturalistic driving studies,” explains Punzo. “This allows us to characterise both the heterogeneity of drivers and driving behaviour in realistic traffic conditions.”
Safer, more sustainable transport
Once the project is complete, its models will immediately be able to serve as a human baseline for defining the required safety level of automated driving systems. Ultimately, the same models will support the automotive industry, its research partners, certification bodies and consumer testing organisations to realistically simulate the behaviour of other human-driven vehicles in mixed traffic. “The i4Driving models will serve as a reference for the design of human-like and therefore easily predictable and acceptable behaviour of automated vehicles in mixed traffic, enabling safe and trustworthy interaction with conventional vehicles,” concludes Rodrigues. “The net result will be a safer and more sustainable transport system and a better use of road infrastructure capacity.”
Keywords
i4Driving, driving behaviour, automated, transportation science, transport and mobility, virtual simulations, automated driving systems, ADS, road safety, driving simulator