Ljubomir Jovanov, Senior Researcher at imec-IPI-Ghent University, will present "AI-Based traffic analytics for improved safety in smart cities" at the AI4 Smart Mobility session during Trefdag Digitaal Vlaanderen (Digital Flanders Meeting Day).
The talk is part of the AI4Growth session around AI in Mobility, planned for 10.45-11.35 am on Thursday, November 25, 2021.
Intelligent technical systems increasingly take up on making decisions that directly influence our lives. For example, today's advanced driver assistance systems (ADAS) make decisions on emergency braking to save lives of pedestrians. We are working in the European project NextPerception to develop sensor fusion technology that will secure better safety of all and especially vulnerable road users (pedestrians, cyclists).
Firstly, we are developing advanced information fusion that relies on information from heterogeneous sensors. The first level of information fusion is applied on data acquired from the sensors at the same location. Subsequently we fuse the information at the one intersection, and information from multiple intersections. For this purpose, we have developed collaborative fusion methods. The second challenge we are working on is optimal communication between resources at different levels. An important aspect of our work is distributed computation and communication between computational levels. Finally, we propose a federated learning approach to provide continuous learning of AI models to provide better adaptation to a situation and more efficient use of resources. Such an approach leverages communication requirements and allows decentralization and the use of low-cost hardware.