MASTER THESIS: EVENT-BASED NEUROMORPHIC VISION FOR ROBOTICS (M/W/D)

You will be integrated in a small team of neuromorphic computing researchers at fortiss. You will be working on one of our neuromorphic vision sensors:

• an active object identification sensor based on infrared emitters, an event camera and neuromorphic processing
• a body motion recognition sesor with an event-based camera and online few-shots learning on neuromorphic hardware (Intel Loihi)

This works, which build up on prior simulation prototypes, are part of projects with very large industrial partners. In either topic, you will learn about existing simulated sensors and implement a hardware prototype running code on neuromorphic hardware on real use cases, such as asset monitoring or robot interaction. You will optimize the vision algorithm to suit real conditions and close the sim2real gap. In a second phase, you will enhance the efficiency of the prototype sensor and extend it to a broader scope such as recognize multiple movements or multiple actors at once and estimate their movements amplitude. A publication, in a scientific journal or conference, as a co-author of your supervisor, is most likely expected at the end of your mission.