PhD - Gent | More than two weeks ago
Many smart mobility and industrial automation applications rely on object detection and tracking. Examples include autonomous vehicles, smart traffic infrastructure, mobile robots, security & surveillance. Sensor fusion is essential for safety critical systems, both for robustness (redundant information) and accuracy (complementary information). Standard sensor fusion techniques are unable to cope with difficult circumstances, e.g. partial blindness of a sensor due to adverse weather or light conditions, or objects of interest partly occluded by static scene elements. In this research, we will improve upon the state of the art for fusion of cameras (RGB cameras, thermal cameras, DVS) with radar and/or lidar, by accurately modeling their behavior under different complicating circumstances, and by building a probabilistic framework to assess the reliability of the fused output. Additionally, novel forms of sensor fusion will be explored that include feedback loops and low-level communication between the sensors. These novel concepts require detailed theoretical modelling of the data dependencies to ensure both maximum performance and graceful degradation in difficult conditions.
This topic builds on extensive prior experience within IPI regarding sensor fusion for autonomous driving, traffic monitoring and industrial safety.
Required background: Engineering or Mathematics, with strong interest in Computer Science
Type of work: 45% theoretical modeling, 45% algorithm implementation, 10% field testing
Supervisor: Hiep Luong
Co-supervisor: Wilfried Philips
Daily advisor: David Van Hamme
The reference code for this position is 2022-111. Mention this reference code on your application form.