Integrated Information Processing
Integrated Information Processing Group
The Integrated Information Processing (IIP) Group carries out research in the following areas:
The main focus of the IIP Group is on theory, algorithm design, and hardware implementation of new technologies for beyond fifth-generation (5G) wireless communication systems. Most projects focus on emerging communication technologies including massive MIMO, millimeter-wave (mmWave) and terahertz communication, cell-free massive MIMO, intelligent reflective surfaces, ultra low-latency short-packet transmission, and testbed design for massive MIMO prototyping.
Indoor positioning and outdoor positioning in urban scenarios of mobile phones is a notoriously difficult task. Recently, tools from machine learning have been used to perform positioning from channel-state information (CSI). Most projects focus on channel charting, a new technology developed in the IIP group that enables self-supervised positioning from CSI without the users' consent.
Modern wireless systems are equipped with large arrays of parallel radio-frequency (RF) chains. Such RF chains are extremely accurate sensors that can be used not only for high-rate data transmission but also for sensing. Most projects in the realm of the emerging paradigm of simultaneous sensing and communication (SISCO) are on imaging the area next to the antenna array and on classification of user behavior using machine learning techniques.
Processing in memory (PIM) moves computation into memories with the goal of improving throughput and energy-efficiency compared to traditional von Neumann-based architectures. Most projects in this research area are in designing all-digital and semi-custom PIM accelerators that can be fabricated with conventional CMOS technologies and for emerging applications in machine learning, signal processing, and wireless communication.
Always-on sensors that continuously monitor the environment for certain events must operate with energy-efficient classification and detection pipelines. The projects in this area build upon a novel classification pipeline developed in the IIP group called analog-to-feature (A2F) conversion that directly acquires features in the analog domain using non-uniform wavelet sampling (NUWS). Possible applications are real-time sensing and classification of EEG, ECG, RF, and audio signals.