Following the successful ‘Forum on Embedded AI’ of D3A 1.0, this session targets academic scholars and industrial researchers with an interest in Embedded AI and TinyML. The purpose is to reflect on the presented ideas, exchange knowledge between the participants and explore potential collaborations.

The session will cover talks around Artificial Intelligence (AI) and how it is utilized and applied within the Cyber-Physical Systems (CPS) domain. Deploying AI closer to the edge, where the data is generated, creates new opportunities, but also new challenges due to the limited resources of CPS. After few years that AI has been used in CPS, Embedded AI and TinyML have become more mature and we aspire that this session will serve as a forum to disseminate new results to the community but also present current challenges which will lead to future projects.

Program

Opening introduction (2m)
Session organiser(s)

Keynote presentation: Towards Efficient and Flexible Training and Inference on the Edge (40m)
Olaf Landsiedel (Kiel, DE), keynote speaker

Wake Vision Dataset (12m)
Emil Njor (DTU)

TinyML to Monitor a Colossal Industrial CPS (12m)
Iman Sharifirad (AU)

Towards Intermittent Neuromorphic Computing (12m)
Junaid Qazi (DTU)

Heterogeneous Sub-8-Bit Quantization of Neural Networks (12m)
Luca Pezzarossa (DTU)

Organizers

• Charalampos Orfanidis (chaorf@dtu.dk), Postdoc, DTU (Lead Organiser)
• Emil Jørgensen Njor (emjn@dtu.dk), PhD Student, DTU
• Jalil Boudjadar (jalil@ece.au.dk), Associate Professor, AU
• Luca Pezzarossa (lpez@dtu.dk), Associate Professor, DTU
• Xenofon Fafoutis (xefa@dtu.dk), Professor and DIREC WS6 Manager, DTU

Level

Intermediate