Multi-cue perception supported by background knowledge ((09) R02)

• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: since 2017
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 329551904

The goal of this subproject is to design a perception module that interprets the different sources of sensor data, i.e. subsymbolic information, by utilizing background knowledge and memory of past events, including symbolic information, to support this interpretation process. In Phase II, we want to extend this successful two-stage design of Deep Learning plus geometric Bayesian fusion to go beyond pre-known rigid objects while staying in a finite dimensional parameter space.

• DFG Programme: Collaborative Research Centres
• Subproject of: SFB 1320: EASE – Everyday Activity Science and Engineering
• Applicant Institution: Universität Bremen
• Project Head: Professor Dr.-Ing. Udo Frese