senior development engineer, francesco rossetto, presents two different setups for testing ai/ml csi compression, both using the r&s®cmx500 base station emulator. join us for a detailed look into testing this exciting new feature.
ai/ml can improve the csi report, a fundamental part of a cellular system. in ai-based csi report compression, the neural network compresses or decompresses a set of beamformers. the two models – compressor and beamformer – are deployed at two sides of a communication link: the base station and the user equipment (ue). testing this feature requires (1) a base station emulator that can run a neural network in real time and (2) a ue capable of csi compression. however, this feature is not yet officially standardized in 5g, so the ue must be replaced with a box that can emulate ai/ml csi compression.
