A human driver constantly makes decisions while driving. Such decisions concern, for example, how to merge into traffic from a highway ramp, whether and how to overtake a slower vehicle, how to negotiate a crossing, or how to select the appropriate speed in a residential area. The decision process is very complex, and involves potentially conflicting demands when balancing safety against efficiency.
There are always uncertainties in traffic situations. Sensors are noisy and do not capture the complete traffic situation (for example, due to occlusions, limited sensor range, limited updating frequency etc.). Moreover, the intent and future actions of other vehicles will not be
known, especially if those vehicles are manually driven. In order to generate a fully autonomous vehicle, one must develop a
decision-making system capable of handling all such uncertainties, while balancing safety versus efficiency in an optimal way. How to do
this is still an open research question.
This project will address traffic decision-making from the perspective of machine learning. The project will focus on a time horizon of 1 to 10s. Thus, it will neither involve more long-term route planning nor actuator coordination (to achieve different movements) which is carried out on a faster time scale.
Associate Professor at Chalmers, Mechanics and Maritime Sciences, Vehicle Engineering and Autonomous Systems
Doctoral Student at Chalmers, Mechanics and Maritime Sciences, Vehicle Engineering and Autonomous Systems
Adjunct Professor at Chalmers, Mechanics and Maritime Sciences
Full Professor at Chalmers, Mechanics and Maritime Sciences, Vehicle Engineering and Autonomous Systems
Funding Chalmers participation during 2020–2021
Areas of Advance
Areas of Advance