Performance optimization for cognitive radio networks with RF energy harvesting capability
Dinh, Thai Hoang
Date of Issue2016-05-10
School of Computer Engineering
A cognitive radio is an intelligent radio network in which unlicensed users (i.e., secondary users) can opportunistically access idle channels when such channels are not occupied by licensed users (i.e., primary users). The main purpose of cognitive radio networks (CRNs) is to utilize the available spectra which become more and more scarce due to the boom of wireless communication systems. In CRNs, wireless nodes can operate based on an AC power supply. However, this is feasible only for fixed wireless nodes. A mobile node instead has to operate on the energy from a battery. Accordingly, the battery has to be charged or replaced regularly to ensure sufficient energy supply for the mobile nodes. Alternatively, energy harvesting techniques have been introduced as a promising solution to perpetuate operations of the mobile nodes. With wireless energy transfer, batteries can be replenished without using any physical connection for charging or battery replacement. Recently, radio frequency (RF) energy harvesting techniques with high efficiency have been introduced. Such techniques allow a wireless node to harvest and convert electromagnetic waves from ambient RF sources (e.g., TV, radio towers and cellular base stations) into energy which can be used for data transmissions. With the RF energy harvesting capability, the wireless node can continue its operation without physically changing or recharging its battery. In RF powered CRNs, secondary users can harvest energy from radio signals by using RF energy harvesting devices and then use such energy to transmit data to the primary channels. To obtain enough energy and spectrum access opportunities for data transmissions, the secondary users must search for not only idle channels to transmit their packets, but also busy channels to harvest RF energy. The channel access, which determines the channel to transmit packets or to harvest RF energy, is a crucial step to achieve optimal performance for the secondary system. Therefore, in this thesis, we mainly focus on channel access strategies in RF powered CRNs with the goal of optimizing for the secondary network performance while still guaranteeing the quality-of-service for primary systems. The thesis presents two major contributions. Firstly, we study the channel access problem in which a secondary user selects a channel to access for packet transmissions or to harvest RF energy. We formulate the channel access problem as a Markov decision process (MDP) and adopt the linear programming technique to maximize the throughput for the secondary user. We then propose an online learning algorithm to obtain the optimal channel access policy for the secondary user. With the proposed learning algorithm, the secondary user can observe and adapt the channel access decision to achieve its objective. Secondly, we study the scenario with multiple secondary users coexisting in the same network and they want to cooperate to maximize the joint objective. In this case, we introduce an approach using a round robin scheduling and each secondary user is equipped with an online learning algorithm in order to explore the surrounding environment and make optimal decisions. After that, we introduce a decentralized partially observable MDP framework to model the cooperation among secondary users without coordination and propose a decentralized learning algorithm through using the policy gradient and the Lagrange multiplier method. In summary, this thesis mainly focuses on the channel access problem for RF powered CRNs which is one of the most important issues of CRNs as well as wireless powered CRNs. In all research works studying channel access strategies in wireless powered CRNs in the literature, they considered only one licensed channel. To the best of our knowledge, this is the first work that studies the performance optimization problem for RF powered CRNs with multiple licensed channels taken into considerations. In addition, in this thesis, we also present novel models as well as solutions to address the channel access problem.