Multi-Receiver Automatic Modualtion Recognition

Researching solutions to advance the state-of-the-art in Automatic Modulation Recognition (AMR) for radio communications. We created a new dataset for AMR with muliple receivers observing the same transmited signal propogated through a simulated generated topography. We modified existing state-of-the-art models to process the multi-channel nature of the dataset. We then deployed the model on a Xilinx RFSOC and a Jetson Nano to demonstrate the feasibility of deploying multi-channel AMR on edge devices. Models trained on the dataset outperformed models trained on exisitng single-receiver datasets. We show that muliple receivers provides significant classification performance imporvements at the same receive power level with accuracy mroe than doubling at 0dB and -10dB.