New approaches for quality of service provisioning in cognitive radio networks

Along with the proliferation of wireless communication with diverse services and applications, spectrum efficiency has become a major concern in both licensed and unlicensed frequency bands. Cognitive radio networks (CRNs) stand as a solution for meeting the increasing demand on physically limited frequency spectrum by defining the roles of secondary users (SUs) and secondary networks. These users can adopt their radio transmissions with a dynamic and opportunistic manner with respect to the licensed primary users (PUs) using their software controllable and multi-interface transceivers. Even if the cognitive radio technology defines constrained and secondary networks, along with the increased spectrum demands, the overall network traffic volume and the quality of service (QoS) requirements of associated SUs is expected to increase substantially. In this thesis, encouraged by this forecast, new approaches for QoS provisioning over CRNs are proposed and evaluated. In the first part of the thesis, detailed in Chapter 3, a channel assignment algorithm is proposed for SUs with traffic demands at heterogeneous QoS classes, that is advantageous over the native schemes. The proposed channel assignment algorithm uses an estimation on the PU activity temporal level in the SU packet selection. The temporal PU activity level is estimated by the distributed SUs that only use their local sensing histories via projecting the moving average value of their individual sensing histories into three temporal activity classes. The devised heuristic cognitive channel selection algorithm is first evaluated for preliminary results for a single SU and a single channel environment. Afterwards, considering a more realistic scenario, multi-channel CRN model is studied. The proposed channel selection method is expanded into a multi-channel case. When the designed model is expanded to have multi-user property, a distributed scheduling mechanism is also proposed to alleviate the problem of managing conflicting interests of distributed and selfish SUs within a single collision domain. The proposed scheduling mechanism is a fast and a distributed scheduling method that only uses a centralized information broadcast upon a topology change (i.e. node addition or removal from the system). Both spectrum assignment and scheduling algorithms were evaluated using MATLAB models and compared with the native spectrum assignment scheme and random scheduling in several environments. The simulation results show that the proposed method is advantageous in terms of the Figure of Merit (FoM) metric defined by the weighted sum of throughput ratios of the packets successfully reached to the destination. The proposed method, which basically is a fast PU temporal activity level prediction and a channel selection scheme, reflects its advantageous side especially when the spectrum holes belong to a heavily used PU channel. Later in the second part of the thesis, as a focused case study, MPEG-2 based encoded video transmission problem is studied in Chapter 4 where a distributed and frame-type-differentiated queue management scheme is proposed and evaluated. The proposed queue management method changes the transmit order of the frames in favor of the ones needed by others for them to be decoded and basically decreases the packet loss probability by sending the reference frames first, in order to prevent indirect frame loss due to the decoding dependencies. The discrimination on the packet selection takes place among different video frames within the encoded video burst. The performance of the proposed scheme is investigated with the developed MATLAB model that is supported by FFmpeg tool for encoding operations. The results validate that the proposed scheme outperformed the native scheme that does not discriminate any internal frame type information among the video packets, in various simulation setups in terms of several video quality metrics, namely: Peak Signal to Noise Ratio (PSNR), Mean Structural SIMilarity metric MSSIM), and decodable frame ratio (Q). The contributions of this thesis are the solutions presented for the efficient spectrum management and quality of service provisioning for multimedia traffic transmission in CRNs. Thesis studies cover several CRN models and novel channel assignment and scheduling strategies that are advantageous over native schemes.