Modelling Interaction between Cyclists and Automobiles - Final Report

Report, 2020

The MICA project modelled driver behaviour, focusing on the approaching phase of an overtaking manoeuvre, when a driver moved toward a cyclist while facing oncoming traffic (Euro NCAP test protocols inspired this scenario.). The model predicts the probability for drivers to brake or steer as they approach the cyclists to perform an accelerative (overtake after the oncoming traffic has passed) or flying (overtake before the oncoming traffic has passed) manoeuvre, respectively. This model has been integrated into a smart collision-avoidance system, that provides early (and yet acceptable) warnings and interventions. A virtual assessment estimated the safety benefits of the smart collision-avoidance system using UDRIVE naturalistic data. Our analyses show that the new smart collision-avoidance system can significantly reduce fatalities and severe injuries when compared to traditional collision-avoidance systems, with the new collision warning alone promising a reduction of fatalities by 53-96% and a reduction of serious injuries by 43-93%. This work has been carried out by three PhD students and is now continuing in the MICA2 project.

 

The main deliverables of the project were:

1)      a unique dataset collected on the airfield in Vårgårda where participants interacted with two robots,

2)      a new modelling framework that helps to identify interaction on a scenario basis,

3)      a novel driver model, which can predict overtaking strategy in real-time,

4)      a smart collision-avoidance system which uses the driver model to generate warnings and automated interventions, and

5)      a safety benefit analysis, proving the potential for the new collision-avoidance systems to save lives and reduce injuries from naturalistic European data.

 

Nine scientific contributions describe MICA’s results: one licentiate thesis, two podium presentations to the International Cycling Safety Conference (2018 and 2019, respectively), one conference paper submitted to the Transport Research Arena 2020, and five journal papers.

 

MICA highlighted that:

1)      Modelling the interaction between the overtaking vehicle and the oncoming vehicle is an essential step to increase overtaking safety.

2)      The approaching phase of an overtaking manoeuvre is not necessarily the riskiest; the most significant margin for improving safety may lay in developing systems that support the drivers in the returning phase.

3)      In the approaching phase of an overtaking manoeuvre, the potential safety benefits from automated emergency steering (a system not addressed in MICA) is substantial.

4)      As an overtaking manoeuvre develops from the approaching to the steering, passing, and returning phase, vehicle kinematics and proximities become more critical, challenging active safety systems and calling for new passive safety solutions.

5)      More experimental data, collected in more critical situations than what was possible in MICA, is needed to address overtaking safety properly. New methodologies, such as augmented reality and virtual reality, offer the best opportunities to collect such data without ethical concerns.

6)      More naturalistic data is needed to validate our driver models and the new systems that we started developing in MICA.

7)      Interaction among road users is complex and models of vulnerable road-user behaviour are also needed to make robust predictions. As we move from an overtaking scenario to a crossing scenario, this aspect will become even more crucial.

 

MICA2, a new FFI project including Volvo Cars, Autoliv, Veoneer, Viscando, if, VTI, and Chalmers, will now address these issues.