The research areas of the subproject are environmental detection, prediction, situation assessment, and research and development of communication solutions.
With the proliferation of self-propelled vehicles and robots, there is a growing need to develop automated features for environmental detection and navigation of off-road vehicles, for both civilian and military applications. Current state-of-the-art self-driving solutions (e.g. Tesla, Waymo) are increasingly approaching people’s driving skills on a highway or in a known enclosed area (e.g. a site) and we are witnessing rapid development in well-defined urban environments where vehicles and pedestrians moving on good quality, regularly painted roads, and traffic dynamic controlled by traffic signs and traffic lights can be expected. However, these solutions are not yet prepared for robust operation in a completely unknown environment with changing terrain conditions. These include abandoned “semi-urban” areas on the outskirts of the city, dirt roads or completely offroad transport.
In autonomous transport, an important task is to constantly “monitor” other traffic participants, predict expected movements and find out where the people involved are paying attention, how they might react in a possible dubious transport situation, and how this can be related to their current "Region of Interest".
Tasks:
Research on behavioral prediction, maneuver detection using machine learning methods. Teaching and classifying convolutional neural networks and encoder and decoder structures.
Application of multi-sensor and multi-object filters to eliminate measurement errors and track objects. Experiments using multiple motion models