Reconfigurable Digital Channelizer Design Using Factored Markov Decision Processes

Abstract

In this work, a novel digital channelizer design is developed through the use of a compact, system-level modeling approach. The model efficiently captures key properties of a digital channelizer system and its time-varying operation. The model applies powerful Markov Decision Process (MDP) techniques in new ways for design optimization of reconfigurable channelization processing. The result is a promising methodology for design and implementation of digital channelizers that adapt dynamically to changing use cases and stochastic environments while optimizing simultaneously for multiple conflicting performance goals. The method is used to employ an MDP to generate a runtime reconfiguration policy for a time-varying environment. Through extensive simulations, the robustness of the adaptation is demonstrated in comparison with the prior state of the art.

Runtime Adaptation in Wireless Sensor Nodes Using Structured Learning

“…One of the earliest known applications of using an MDP to control resources in computing systems at runtime is [12]. Other notable examples of differing approaches in the literature include a reconfigurable router [13], a reconfigurable digital filter bank [14,15], a power management module for a microprocessor [16], and a smartphone scheduling program that synchronizes email efficiently [17].…”

“…Some of these CMM techniques seek to reduce the storage size of the MDP's data structures by exploiting some structural component embedded within the MDP (e.g., see [2,25,26,27,28]). Other techniques involve modeling approaches that reduce the MDP state space via generalization and abstraction of system dynamics (e.g., see [14,15]). Another approach has been to keep algorithms and data structures as is, and take advantage of recent advancements in parallel processing using embedded GPUs, for example [29,2].…”

“…In [14], we introduced the concept of transition states, which is based on a similar concept initially proposed in [12]. We expanded and elaborated on the initial concept in both [14,15].…”

“…The only transition probability needed is derived from an estimate of the expected time until the system leaves the set of states. A discussion on how to apply this to stochastic dynamics, and some of the trade-offs associated with this concept can be found in [14,15].…”

“…Following the Transition States theory from the literature (e.g. [15]), the value of the parameter ρ is defined in Equation 16, where T F is the duration of one control frame.…”