Embedded AI/ML is seen as a key technology to enable future 6G wireless communications systems. Currently there is a lot of research ongoing to embed AI/ML in almost every layer of future communications system. The vision is to build communications systems which can easily learn to adapt to the exploding number of deployment and application scenarios by just providing them with the right training data. The latter is one of the big challenges of designing trustworthy AI/ML-based communication systems. To ensure that they work robustly and reliably in all the different real-world deployment scenarios, we need to train and validate these systems with data in sufficient quality and quantity. Real-world Radio Frequency (RF) data play an important role in that process.
In our presentation we will present NI’s RF data recording system with an open-source Python-based API that allows you to easily setup and automate expansive RF data recording campaigns in complex RF environments. The recorded data sets are stored in the open-source Signal Metadata Format (SigMF) with a comprehensive and standardized scenario description provided as metadata. This makes them broadly usable for training and validation of wireless AI/ML applications as we will show for a relevant RF spectrum sensing application.