Enhancing Guessing Random Additive Noise Decoding using Channel Estimation

We investigate optimal channel estimation strategies for Guessing Random Additive Noise Decoding (GRAND). We explore the trade-off between channel sounding and block error rate (BLER) performance using empirical techniques. We show that correlation in the noise introduced by the uncertainty in the channel estimate can be exploited to improve BLER performance. Using a soft-information variant of the GRAND algorithm called Ordered Reliability Bit Guessing Random Additive Noise Decoding - Approximate Independence (ORBGRAND-AI), we are able to improve decoding performance by considering the correlation over small blocks of symbols. We compare ORBGRAND-AI decoding to interleaved schemes which have been stripped of correlation and find that up to 0.5 dB gain can be achieved.