A Variable-Order Discrete Model for the Fading Channel

Fading on the mobile propagation channel is modeled as a Markov chain with unspecified state space and order. State space and memory are selected via the context tree pruning (CTP) algorithm that has provable optimality in fitting Markov models of unknown order. CTP provides a hierarchy of models of increasing dimension each of which may be considered the “best” approximation of the fading channel among Markov models of that dimension. A particular model may then be selected by a criterion appropriate for the task that the discrete model will perform. An example herein considers the model’s ability to faithfully mimic statistics of sojourn times in each channel state. That criterion choice is relevant for simulation of channel access and use schemes that exploit channel memory.