Balancing forward and feedback error correction for erasure channels with unreliable feedback

The traditional information theoretic approach to studying feedback is to consider ideal instantaneous high-rate feedback of the channel outputs to the encoder. This was acceptable in classical work because the results were negative: Shannon pointed out that even perfect feedback often does not improve capacity and in the context of symmetric DMCs, Dobrushin showed that it does not improve the fixed block-coding error exponents in the interesting high rate regime. However, it has recently been shown that perfect feedback does allow great improvements in the asymptotic tradeoff between end-to-end delay and probability of error, even for symmetric channels at high rate. Since gains are claimed with ideal instantaneous feedback, it is natural to wonder whether these improvements remain if the feedback is unreliable or otherwise limited. Here, packet-erasure channels are considered on both the forward and feedback links. First, the feedback channel is considered as a given and a strategy is given to balance forward and feedback error correction in the suitable information-theoretic limit of long end-to-end delays. At high enough rates, perfect-feedback performance is asymptotically attainable despite having only unreliable feedback! Second, the results are interpreted in the zerosum case of “half-duplex” nodes where the allocation of bandwidth or time to the feedback channel comes at the direct expense of the forward channel. It turns out that even here, feedback is worthwhile since dramatically lower asymptotic delays are possible by appropriately balancing forward and feedback error correction. The results easily generalize to channels with strictly positive zero-undeclared-error capacities.