Traffic Pattern Prediction Based Spectrum Sharing for Cognitive Radios [chapter]

Xiukui Li, Seyed A. Reza Zekavat
2009 Cognitive Radio Systems  
Traffic Model Traffic data patterns can be classified as (Haykin, 2005) : 1) deterministic patterns: for example, each primary user (e.g., TV transmitter) is assigned a fixed time slot for transmission, and when it is switched off, the frequency band is vacated; 2) stochastic patterns: for example, the arrival times of data packets are modeled as a Poisson process, while the service times are modeled as exponentially distributed, depending on whether the data are of packet-switched or
more » ... itched. Note that, in common channel signaling network, the exponential distribution drastically underestimates the proportion of short calls that do not last longer than the mean holding time (Bolotin, 1994) . In general, the traffic stochastic parameters vary slowly. Hence, they can be estimated using historical data. The traffic model built on historical data www.intechopen.com %QIPKVKXG 4CFKQ 5[UVGOU enables secondary users to predict the future traffic pattern of primary users (Li & Zekavat, 2008 ). An overview of traffic modeling is provided in (Frost & Melamed, 1994) . It discusses Markov modulated traffic models, autoregressive traffic models and self-similar traffic models, etc. In addition, the authors in (Adas, 1997) discuss different traffic models in telecommunication networks. Traffic Model for Voice Communication A speech process can be modeled as a two-state Markov chain which alternates between talk spurt and silent periods (Li, 1990) ; (Gruber, 1981) . The authors in (Hong & Rappaport, 1986 ) propose a traffic model for cellular mobile radio telephone systems. The basic system model in (Hong & Rappaport, 1986) assumes that the new call origination rate is uniformly distributed over the mobile service area, and the channel holding time is approximated to a negative exponential distribution. Considering a one-dimensional mobile system with cells in series (e.g., in highways), the authors in (Pavlidou, 1994) uses two-dimensional state diagrams to analyze the traffic in the mixed media cellular system. It assumes four Poisson arrival streams are entering each cell, which are originating new voice calls, originating new data packets, hand-off voice calls and hand-off data packets. The authors in (Leung et al., 1994) also consider the circumstance of a one-way, semi-infinite highway. With the assumption that there are an infinite number of channels available, they present a deterministic fluid model and a stochastic traffic model for a wireless network along the highway. Traffic Model for Video Data The authors in (Dawood & Ghanbari, 1999) provide a summary for traffic models of video data. (Lucantoni et al., 1994) proposes to model a single video source as a Markov renewal process whose states represent different bit rates. Some other models including Markov Modulated Fluid Flow (MMFF) model (Maglaris et al., 1988) , Markov Modulated Poisson Process (MMPP) (Skelly, 1993) , and AutoRegressive AR(1) stochastic model (Maglaris et al., 1988) are proposed to address the basic characteristics of the variable bit rate traffic. The MMFF and MMPP are suitable for queueing analysis of packet switched networks. The AR(1) stochastic model primarily characterizes the inter-frame source bit-rate variations and correlation. For variable bit-rate traffic, the authors in (Knightly & Zhang, 1997) introduce a new deterministic traffic model called deterministic bounding interval-length dependent (D-BIND) to capture the multiplexing properties of bursty streams. For a large-scale satellite network simulation, (Ryu, 1999) proposes: 1) a discrete autoregressive process for MBone ("multicast backbone") video source modeling; 2) the superposition of fractal renewal processes (Sup-FRP) model for Web request arrivals, and, 3) a generalized shot-noise-driven Poisson point process (GSNDP) for aggregate traffic flow modeling. Traffic Model for General Packet Data Based on an analysis of Internet protocols for data communication, (Anderlind & Zander, 1997 ) proposes a simple model for future data traffic in wireless radio networks. Model parameters are selected to describe traffic from the Worldwide Web (WWW) access and from distributed file systems. A multilayer Markov model is considered in (Filipiak, 1992) for arrivals of calls, bursts, and packets to fast packet switching system, where the multilayer refers to call layer, packet layer and burst layer. www.intechopen.com 6TCHſE 2CVVGTP 2TGFKEVKQP $CUGF 5RGEVTWO 5JCTKPI HQT %QIPKVKXG 4CFKQU
doi:10.5772/7838 fatcat:jxcxkyt2mvgchg6mrs6amt2iby