An evasive manoeuver capability to be used in case of on–road emergencies was developed at the Budapest University of Technology and Economics (BME). The manoeuvre planning makes use of a machine learning approach. The evasive manoeuver capability that had been tested only on model cars previously was transferred to a real road vehicle. The capability was successfully validated and demonstrated on the ZalaZone test track for this real vehicle.
The track used for the validation was designated in accordance with the relevant civil engineering standards. The road vehicle used for the validation performed a manoeuver that requires high precision both in timing and in path-following. This dual precision was achieved via a combination of machine learning approaches and of a model predictive control design.
The demonstrated capability stems from the research and development collaboration between the Machine Learning and the Automated Vehicle Control research groups.
The results of this joint effort are summarized in the video appearing below.