Improving the efficiency of spectrum access and utilization under the umbrella of cognitive radio (CR) is one of the most crucial research areas for nearly two decades. The results have been algorithms called cognitive radio engines which use machine learning (ML) to learn and adapt the communication's link based on the operating scenarios. While a number of algorithms for cognitive engine design have been identified, it is widely understood that significant room remains to grow the capabilities of the cognitive engines, and substantially better spectrum utilization and higher throughput can be achieved if cognitive engines are improved. This requires working through some difficult challenges and takes an innovative look at the problem. A tenet of the existing cognitive engine designs is that they are usually designed around one primary ML algorithm or framework. In this dissertation, we discover that it is entirely possible for an algorithm to perform better in one operating scenario (combination of channel conditions, available energy, and operational objectives such as max throughput, and max energy efficiency) while performing less effectively in other operating scenarios. This arises due to the unique behavior of an individual ML algorithm regardless of its operating conditions. Therefore, there is no individual algorithm or parameter sets that have superiority in performance over all other algorithms or parameter sets in all operating scenarios. Using the same algorithm at all times may present a performance that is acceptable, yet may not be the best possible performance under all operating scenarios we are faced with over time. Ideally, the system should be able to adapt its behavior by switching between various ML algorithms or adjusting the operating ML algorithm for the prevailing operating conditions and goal.
In this dissertation, we introduce a novel architecture for cognitive radio engines, with the goal of better cognitive engines for improved link adaptation in order to enhance spectrum utilization. This architecture is capable of meta-reasoning and metacognition and the algorithms developed based on this architecture are called metacognitive engines (meta-CE). Meta-reasoning and metacognitive abilities provide for self-assessment, self-awareness, and inherent use and adaptation of multiple methods for link adaptation and utilization. In this work, we provide four different implementation instances of the proposed meta-CE architecture. First, a meta-CE which is equipped with a classification algorithm to find the most appropriate individual cognitive engine algorithm for each operating scenario. The meta-CE switches between the individual cognitive engine algorithms to decrease the training period of the learning algorithms and not only find the most optimal communication configuration in the fastest possible time but also provide the acceptable performance during its training period. Second, we provide different knowledge indicators for estimating the experience level of cognitive engine algorithms. We introduce a meta-CE equipped with these knowledge indicators extracted from metacognitive knowledge component. This meta-CE adjusts the exploration factors of learning algorithms to gain higher performance and decrease training time. The third implementation of meta-CE is based on the robust training algorithm (RoTA) which switches and adjusts the individual cognitive engine algorithms to guarantee a minimum performance level during the training phase. This meta-CE is also equipped with forgetting factor to deal with non-stationary channel scenarios. The last implementation of meta-CE enables the individual cognitive engine algorithms to handle delayed feedback scenarios. We analyze the impact of delayed feedback on cognitive radio engines' performances in two cases of constant and varying delay. Then we propose two meta-CEs to address the delayed feedback problem in cognitive engine algorithms.
Our experimental results show that the meta-CE approach, when utilized for a CRS engine performed about 20 percent better (total throughput) than the second best performing algorithm, because of its ability to learn about its own learning and adaptation. In effect, the meta-CE is able to deliver about 70% more data than the CE with the fixed exploration rate in the 1000 decision steps. Moreover, the knowledge indicator (KI) autocorrelation plots show that the proposed KIs can predict the performance of the CEs as early as 100 time steps in advance. In non-stationary environments, the proposed RoTA based meta-CE guarantees the minimum required performance of a CRS while it’s searching for the optimal communication configurations. The RoTA based meta-CE delivers at least about 45% more data than the other algorithms in non-stationary scenarios when the channel conditions are often fluctuating. Furthermore, in delayed feedback scenarios, our results show that the proposed meta-CE algorithms are able to mitigate the adverse impact of delay in low latency scenarios and relieve the effects in high latency situations. The proposed algorithms show a minimum of 15% improvement in their performance compared to the other available delayed feedback strategies in literature.
We also empirically tested the algorithms introduced in this dissertation and verified the results therein by designing an over the air (OTA) radio setup. For our experiments, we used GNU Radio and LiquidDSP as free software development toolkits that provide signal processing blocks to implement software-defined radios and signal-processing systems such as modulation, pulse-shaping, frame detection, equalization, and others. We also used two USRP N200 with WBX daughterboards, one as a transmitter and the other as a receiver. In these experiments, we monitored the packet success rate (PSR), throughput, and total data transferred as our key performance indicators (KPI). Then, we tested different proposed meta-CE algorithms in this dissertation to verify the productivity of the proposed algorithms in an OTA real-time radio setup. We showed that the experiments’ outputs support our simulations results as well.
| Identifer | oai:union.ndltd.org:arizona.edu/oai:arizona.openrepository.com:10150/626755 |
| Date | January 2018 |
| Creators | Asadi, Hamed, Asadi, Hamed |
| Contributors | Marefat, Michael M., Bose, Tamal, Marefat, Michael M., Bose, Tamal, Marcellin, Michael W. |
| Publisher | The University of Arizona. |
| Source Sets | University of Arizona |
| Language | en_US |
| Detected Language | English |
| Type | text, Electronic Dissertation |
| Rights | Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author. |
Page generated in 0.003 seconds