Within the 6G paradigm, Integrated Sensing and Communications (ISAC) represents a transformative leap in wireless evolution. As communication networks expand into higher frequency bands (e.g., millimeter-wave and sub-terahertz), signals become vulnerable to dynamic environmental changes. Consequently, environmental sensing becomes critical as it shifts networks from passive data conduits into cognitive infrastructures capable of perceiving physical surroundings, supporting advanced applications like urban digital twins and autonomous vehicular navigation.
The primary challenges of this research lie in extracting environmental monitoring information through light-weight/distributed AI, as well as designing a hybrid AI architecture to intelligently partition and offload complex computational tasks into hardware-constraint platforms. Specifically, the objective is to compress heavy deep learning processes for environmental awareness into ultra-lightweight TinyML modules deployable directly on Software-Defined Radios (SDRs). Seamlessly integrated with Mobile Edge Computing (MEC), this hybridised architecture will enable rapid, low-latency environmental sensing to facilitate the next-gen wireless communication paradigms.
To succeed this project, you will be required to tackle the following/relevant technical bottlenecks:
- Efficiently acquiring and processing complex, high-dimensional environmental sensing data and user physical parameters from both SDR front-ends and complementary sensor networks.
- Architecting hybrid modules to execute high-granularity sensing tasks, e.g. user tracking and identification, dynamic obstacle recognition, and predictive channel estimation, within strict edge resource bounds.
- Robustly decoupling task-specific physical signatures from entangled wireless metrics, including Channel State Information (CSI), amidst severe multipath fading and environmental noises.