Delft University of Technology
Intelligent Vehicles Group
EDL P16-25 P6: Deep Learning for Mobeile Robotics (MRob)
Deep Learning for 3D Semantic Scene Analysis
TMobile robotics require an awareness and understanding of the environments they navigate. In the context of autonomous vehicles, motion prediction of other road users is crucial to inform motion planning and prevent accidents. Current motion prediction methods either rely on hand-crafted rules that generalise poorly or employ black-box models that are difficult to interpret.
The aim of my research is to increase the robustness, interpretability, and efficiency of deep learning approaches in the safety-critical application of autonomous vehicles. A promising research direction to this end is to employ knowledge-driven models in a learnt deep neural network such that the benefits of hand-crafted and pattern-based methods are exploited.