MOTION PREDICTION OF BICYCLISTS IN URBAN ENVIRONMENTS BASED ON LIDAR DATA
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.):¥1,100 Members (tax incl.):¥880
- Publication code
- 20219030
- Paper/Info type
- Other International Conferences
- Pages
- 1-6(Total 6 p)
- Date of publication
- Sep 2021
- Publisher
- JSAE
- Language
- English
Detailed Information
Author(E) | 1) Adrian M. Sonka, 2) Silvia Thal, 3) Roman Henze |
---|---|
Affiliation(E) | 1) Technische Universität Braunschweig, 2) Technische Universität Braunschweig, 3) Technische Universität Braunschweig |
Abstract(E) | Ensuring road safety for bicyclists through an improvement of the foresight of automated vehicles by anticipating critical behavior is a remaining challenge in research and development. A situation adaptive motion planning based on this information can reduce accidents with these vulnerable road users. Our work utilizes machine learning methods for a bicyclist motion prediction, applied to laser scanner data recorded with an own measurement vehicle. Different configurations of multilayer perceptron networks, applied for polynomial coefficient estimation as well as long short-term memory networks are compared and evaluated on a quantitive and qualitative level, surpassing accuracy levels of a physical baseline prediction model. |