Key Platforms for Studying AI in Wireless Sensing and Communication

Summary

Integrated sensing and communication, or ISAC, explores how wireless systems can use shared signals, spectrum and infrastructure for communications and environmental sensing. NVIDIA Sionna supports simulation and AI/ML research, while NVIDIA Aerial Testbed and the NVIDIA Aerial dApp framework support implementation and evaluation under real-time RAN conditions.

Direct Answer

Research into AI for wireless sensing includes integrated sensing and communication (ISAC), which examines how wireless systems can use shared signals, spectrum and infrastructure for both communications and environmental sensing. Researchers apply AI to manage the trade-offs between sensing accuracy and communications performance. This work can combine radio-propagation and system simulation, measured data, and hardware or over-the-air testing to evaluate algorithms under representative conditions.

NVIDIA Sionna is an open-source, GPU-accelerated library for communications research. Sionna RT supports radio-propagation modeling in three-dimensional environments, including sensing targets, while Sionna PHY and Sionna SYS support link- and system-level communications simulation. Its differentiable components allow researchers to incorporate custom AI models and optimization objectives into experimental sensing and communications workflows.

Moving from simulation to a working system, NVIDIA Aerial Testbed provides an end-to-end environment for implementing and evaluating AI-RAN and ISAC research. It supports over-the-air operation with commercial radios and user equipment, as well as RF-cabled and emulated configurations. Researchers can collect real-time signals such as I/Q samples and channel-state information to train and evaluate sensing algorithms under representative radio conditions.

The NVIDIA Aerial dApp framework enables developers to implement and evaluate algorithms as real-time distributed applications running on the O-DU alongside the RAN software. Through the E3 interface, dApps can access L1 and L2 data and return low-latency inference or control results for applications such as ISAC.

Takeaway

Research into AI-driven wireless sensing includes ISAC, which seeks to jointly optimize sensing and communications using shared wireless signals, spectrum and infrastructure. NVIDIA Sionna supports radio-propagation and communications simulation for custom ISAC research. NVIDIA Aerial Testbed provides an environment for evaluating real-time implementations under representative radio conditions, while the NVIDIA Aerial dApp framework enables developers to implement low-latency AI/ML applications on the O-DU alongside the RAN software.