27
May
©geralt - pixabay.com
Improving Motion Prediction in Autonomous Driving with Expert Knowledge - A Bayesian Deep Learning Approach
Nadja Klein, TU Dortmund
27.05.2024
4:30 pm - 6:00 pm
LMU Munich, Ludwigstr. 28 VG/II; Room 211b
Autonomous driving is one of the most highly anticipated yet elusive mobility innovations. The field has made significant advances through deep learning, especially in perception and motion prediction. Still, the field faces open challenges, since safety requirements in autonomous driving demand for robust domain adaptations between locations and well-calibrated uncertainty estimates for the numerous high-risk edge cases in urban environments.
To meet those demands, we explore the potential of Bayesian deep learning methods to motion prediction. We demonstrate how a Bayesian approach can be used to regularize commonly employed motion prediction models by utilizing prior expert knowledge. More specifically, we adapt a CoverNet baseline model with a compute-efficient last layer Gaussian Process approximation and integrate prior drivability knowledge. Doing so improves both, robustness and calibration.
Organized by:
LMU Center for Advanced Management Studies
LMU Munich School of Management
Related
©Bergson
Presentation • 14.10.2024 • Bergson Kunstkraftwerk Studio, Rupert-Bodner-Str. 5, 81245 München
KI - Der Anfang vom Ende oder einfach nur nützlich?
Together with Constanze Zawadzky (Innovation Park AI), MCML PI Reinhard Heckel will give interested parties an insight into the world of artificial intelligence.
©geralt - pixabay.com
Lecture • 16.07.2024 • LMU Munich, Main Building, M 210 und online via Zoom
Nachgefragt – Ethikgespräche an der LMU: Kann ChatGPT denken?
David Lauer from the University of Kiel questions if we're in the age of thinking machines, exploring what "thinking" means and AI's true capabilities and limitations …
©geralt - pixabay.com
Lecture • 16.07.2024 • LMU Munich, Seminar Room 211, 2/F, Ludwigstr. 28 (Front Building)
Methodological Advances in Experimental Design
Talk by Drew Dimmery from the Hertie School Data Science Lab, about improving experimental design for treatment assignment, HTE estimation, and online assignment.
©geralt - pixabay.com
Lecture • 09.07.2024 • LMU Munich, Seminar Room 211, 2/F, Ludwigstr. 28 (Front Building)
Patient Risk Stratification through a Causal Lens
The talk by Jenna Wiens from the University of Michigan examines why many sepsis risk tools fail and suggests using causal reasoning for better outcomes.
©geralt - pixabay.com
Lecture • 04.07.2024 • LMU Munich, Seminar Room 211, 2/F, Ludwigstr. 28 (Front Building)
Large Language Models in the Age of Misinformation
Francesco Pierri from Politecnico di Milano discusses the risks of generative AI, such as GPT-4 and Midjourney, in spreading disinformation and targeting individuals.