Powerline Communication with AI-based Physical Carrier Sensing Providing intelligent decisions within micro-seconds

The powerline communication technology developed at CC ISN combines an OFDM physical layer based on IEEE1901 with a deterministic protocol for multi-master bus arbitration according to ARINC 629. This bus arbitration protocol requires reliable monitoring of the state of the communication channel (channel busy/idle) with a minimum response time. The existing bus monitoring concept is based on time-averaging of signal energies in a number of pre-defined frequency bands. In this project a new bus monitoring concept was developed which uses a RNN trained on time-domain data only. Since no reprocessing is required to transfer data from time- to frequency domain, the model not only provides more deterministic reaction times than the existing approach, but also reacts faster and uses one order of magnitude less hardware ressources. Under a variety of channel and noise conditions the model achieves the same or better detection accuracy than the existing implementation.A real-time fixed-point arithmetic FPGA implementation of the RNN-model has been used to integrate and test the new appoach within the prototyping platform of the PLC modem.