Hardware-aware Neural Networks for Integrated Sensing and Communication

As future communication signals advance in time and angular resolution, Integrated Sensing and Communication (ISAC) emerges as a promising technology for 6G and beyond. ISAC leverages existing communication infrastructure for sensing applications such as generating and maintaining a digital twin of the radio environment to optimize sensing and communication systems to extended reality applications where precise user movement- and gesture-detection are required.

These ISAC applications generate large amounts of data that either have a very high throughput requirement or need to be processed locally in a highly efficient manner.  This processing is often done using neural networks due to their ability to handle large amounts of data. However, many ISAC algorithms remain confined to theoretical models, with a limited focus on hardware-software co-design. Standardized benchmarking frameworks are also absent for the fair evaluation of algorithmic performance across different hardware platforms.

This PhD research will address these gaps by combining hardware-aware neural network mapping expertise on novel AI accelerators with advanced ISAC neural networks. The goal is to evaluate realistic implementations of ISAC on in-house and off-the-shelf hardware. This will provide unique insights into the performance of the algorithms and hardware designed here at imec and provide a benchmarking starting point for the ISAC research community.  You will adapt and create ISAC algorithms and neural network models in a hardware-software co-design approach and map these into an existing communication pipeline. A data set must be collected using a state-of-the-art communication testbed to benchmark your implementations.

The following qu​estions will be investigated:

• How does the hardware mapping of ISAC algorithms influence performance?
• How can ISAC implementations be fairly compared and benchmarked against state-of-the-art solutions on different hardware platforms?

This work will close the hardware-software knowledge gap in ISAC systems, providing the research community with a clear path for future innovations in 6G communication and sensing technologies.

The successful PhD candidate will be part of a large IMEC team working on the research, implementation and prototyping of future communications and AI systems working closely together with a research team at KU Leuven: experts in communications systems, processing, embedded hardware and machine learning. This is a unique opportunity to develop innovative, multi-disciplinary technology and shape future wireless and sensing networks. You will publish your research in top-level journals and conferences.