Speaker: Jörg Conradt, KTH Stockholm, Sweden
Title: Low-power embedded event-based vision processing for low-latency robotics
Abstract: Brain inspired information processing in hardware and in software - or “neuromorphic computing” - opens a possible path to real-time, low-energy computation. Today, various neuromorphic computing systems are available as customizable hardware for small or large applications, from single chips to computer server rooms. For efficient use of such novel hardware, we need to rethink computation in terms of event-based or spiking neuromorphic algorithms, instead of traditional sequential (CPU) or parallel (GPU) computing. In this presentation, I will first briefly introduce the concepts of neuromorphic computing, including a quick overview of present and upcoming neuromorphic hardware for sensing and computation. Following, I will show and discuss multiple application examples of event-based sensing and perception, leading a path towards closed-loop actuated robotic systems.
Short bio: Jörg Conradt is Associate Professor at the KTH in Stockholm, Sweden, School of Electrical Engineering and Computer Science. Before joining KTH, he was W1 Professor at the TU München, Germany. He holds a Ph.D. from ETH Zurich, an M.S. in Robotics from the University of Southern California, and a Diploma in Computer Engineering from TU Berlin. He is an IEEE senior member and was the founding director of the Elite Master Program NeuroEngineering at TUM. Research in his group on Neurocomputing Systems (http://neurocomputing.systems) investigates key principles by which information processing in brains works, and applies those to real-world interacting technical systems.
Date and venue: Thursday 6 February 2025, 14:00-15:00, JDB Seminar Room (and Zoom: https://newnham.zoom.us/j/92544958528?pwd=YS9PcGRnbXBOcStBdStNb3E0SHN1UT09)
