Communication with causal CSI and controlled information outage

In this letter, a Rayleigh block-fading (BF) channel, subject to an information outage probability constraint, is considered. The transmitter is assumed to have causal knowledge of the channel state information (CSI), which is exploited to intelligently allocate the power over the blocks (and hence vary the channel mutual information) to minimize the average transmitted power per block for satisfying the outage probability constraint for a given target code-rate. We first show that the optimal solution to this problem can be obtained by solving the reverse problem of minimizing the outage probability for a range of long-term power constraints through repeated uses of dynamic programming (DP), which is nevertheless prohibitively complex. Then, we develop a suboptimal allocation algorithm which still uses DP to exploit the CSI causality but at a much reduced complexity. A performance lower-bound is further derived, which permits us to see that the proposed algorithm is near-optimal, especially in the small outage probability regime. A scheme called equal-outage-probability per block (EOPPB) which compromises the performance further for reducing the complexity is also devised. To compare the methods, we evaluate both analytically and numerically their complexities and performance. The results are finally generalized to multiple-input multiple-output (MIMO) BF channels.