Minimum-Latency FEC Design With Delayed Feedback: Mathematical Modeling and Efficient Algorithms

In this paper, we consider the packet-level forward error correction (FEC) code design, without feedback or with delayed feedback, for achieving the minimum <italic>end-to-end latency</italic>, i.e., the latency between the time that packet is generated at the source and its <italic>in-order delivery</italic> to the application layer of the destination. We first show that the minimum-latency FEC design problem can be modeled as a partially observable Markov decision process (POMDP), and hence the optimal code construction can be obtained by solving the corresponding POMDP. However, solving the POMDP optimally is in general difficult unless its state and action space is very small. To this end, we propose an efficient heuristic algorithm, namely the majority vote policy, for obtaining a high quality approximate solution. We also derive the tight lower and upper bounds of the optimal state values of this POMDP, based on which a more sophisticated <inline-formula> <tex-math notation="LaTeX">$D$ </tex-math></inline-formula>-step search algorithm can be implemented for obtaining near-optimal solutions. The simulation results show that the proposed code designs via solving the POMDP, either with the majority vote policy or the <inline-formula> <tex-math notation="LaTeX">$D$ </tex-math></inline-formula>-step search algorithm, strictly outperform the existing schemes, for both cases, without or with only delayed feedback.