Reliable Adaptive Recoding for Batched Network Coding with Burst-Noise Channels

Network coding provides an efficient approach for multi-hop communication. The message from the source node is encoded into batches of coded packets, and intermediate nodes perform recoding steps to transmit the message to the destination node. Adaptive recoding optimizes the number of recoded packets per batch to enhance network throughput, accounting for fluctuations in packet loss. In this paper, we propose an adaptive recoding scheme in burst-noise channels with unknown channel parameters. We first provide uncertainty quantification for channel parameters using historical data and build a confidence set to cover true channel parameters with high probability. Next, we obtain the optimal recoding policy by solving a robust Markov decision process (MDP) problem, where uncertain parameters belong to the confidence set. The objective of the robust MDP is to optimize the worst-case reward function by considering all possible problem parameters from the confidence set. Experimental results demonstrate that our proposed recoding strategy significantly enhances network communication throughput with burst-noise channels.