Parallel Channel Over Wide Spectrum: Modeling-, Mining-, and Matching-Based Transmission Optimization

We address the matching-based communication system optimization over parallel channels in wide spectrum. The dimension of the logarithm channel link gain vector can be reduced significantly under certain line-of-sight millimeter wave and optical wireless communication scenarios. We design the quantization points in the reduced dimensional space and store the optimal scheduling policy corresponding to each quantization point. We first investigate the covariance of the achievable rates of the parallel channels and the system outage probability. Then, we formulate a system optimization framework via matching the link gains of parallel channels to the quantization points, and obtain the sufficient conditions such that the utility loss of employing the quantization-based solution can be sufficiently small under sufficiently small quantization distortion in the reduced dimensional space. Such sufficient conditions can be satisfied for certain common optimization examples. We also address the matching-based communication system design under noisy channel estimation via providing a probabilistic upper bound on the utility loss in terms of the quantization design and channel estimation pilot. It can be shown that with sufficiently large probability the utility loss due to the quantization can be sufficiently small, if the quantization distortion is sufficiently small and the pilot energy is sufficiently large.