Spectrum Management Framework for Cognitive Radio Networks

CR networks impose unique challenges due to the coexistence with primary networks as well as diverse QoS requirements. Thus, new spectrum management functions are required for CR networks with the following critical design challenges:

• Interference Avoidance: CR network should avoid interference with primary networks.
• QoS Awareness: In order to decide an appropriate spectrum band, CR networks should support QoS-aware communication, considering dynamic and heterogeneous spectrum environment.
• Seamless Communication: CR networks should provide seamless communication regardless of the appearance of the primary users.

Figure 2. Spectrum Management Framework.

In order to address these challenges, we provide a directory for different functionalities required for spectrum management in CR networks. The spectrum management process consists of four major steps:

• Spectrum Sensing: A CR user can only allocate an unused portion of the spectrum. Therefore, the CR user should monitor the available spectrum bands, capture their information, and then detect the spectrum holes.
• Spectrum Decision: Based on the spectrum availability, CR users can allocate a channel. This allocation not only depends on spectrum availability, but it is also determined based on internal (and possibly external) policies.
• Spectrum Sharing: Since there may be multiple CR users trying to access the spectrum, CR network access should be coordinated in order to prevent multiple users colliding in overlapping portions of the spectrum.
• Spectrum Mobility: users are regarded as ¡±visitors¡± to the spectrum. Hence, if the specific portion of the spectrum in use is required by a primary user, the communication needs to be continued in another vacant portion of the spectrum.

The spectrum management framework for CR network communication is illustrated in Fig. 2. It is evident from the significant number of interactions that the spectrum management functions necessitate a cross-layer design approach. Thus, each spectrum management function cooperates with application, transport, routing, medium access and physical layer functionalities with taking into consideration the dynamic nature of the underlying spectrum.


Related work:

I. F. Akyildiz, W. Y. Lee, M.C. Vuran and S. Mohanty, ``NeXt Generation / Dynamic Spectrum Access / Cognitive Radio Wireless Networks: A Survey," Computer Networks Journal (Elsevier), Vol. 50, pp. 2127-2159, September 2006.

I. F. Akyildiz, W. Y. Lee, M.C. Vuran and S. Mohanty, ``A Survey on Spectrum Management in Cognitive Radio Networks," IEEE Communications Magazine, Vol. 46, pp. 40-48, Apr. 2008.
.

---

Spectrum Sensing for Cognitive Radio Networks

A cognitive radio should monitor the available spectrum bands, capture their information, and then detect the spectrum holes. Hence, spectrum sensing is a key enabling technology in cognitive radio networks. In spectrum sensing, the detection accuracy has been considered as the most important factor to determine the performance of cognitive radio networks. However, in reality, RF frontend of CR users cannot differentiate the primary user signals and CR user signals. In case of the energy detection, widely used in spectrum sensing, transmission and sensing cannot be performed at the same time. Thus, during the sensing(observation time), all CR users should stop their transmissions and keep quiet. Due to this hardware restriction, CR users should sense the spectrum periodically with sensing period Ts and observation time ts, as described in Figure 3.




Figure 3. Periodic spectrum sensing strcuture.

However, the periodic spectrum sensing should consider following design issues:

• Interference Avoidance: In the periodic sensing, interference is related to not only sensing accuracy depending on observation time but also the CR transmission time and tarffic statistics.
• Spectrum Efficiency: The main objective of cognitive radio is the efficient use of spectrum resources. However, since CR users cannot not transmit during the sensing, spectrum efficiency will be degraded in evitably.

In Cognitive radio network, available spectrums may show different characteristics with the bandwidth, the primary user activity, and acceptable interference limit, which affect both the sensing accuracy and spectrum efficiency. Thus, spectral efficient sensing technique is essential for cognitive radio networks. Hence, in this project we will propose the spectral efficient sensing technique for cognitive radio networks, which provides optimal spectrum sensing period and observation time to maximize the efficiency of each spectrum bands subject to the resource limitation and interference restriction.


Related work:

W. Y. Lee, and I. F. Akyildiz, ``Optimal Spectrum Sensing Framework for Cognitive Radio Networks," To appear in IEEE Transaction on Wireless Communications, 2008.

---

Spectrum Decision Framework for Cognitive Radio Networks

In cognitive radio (CR) networks, unused spectrum bands will be spread over a wide frequency range including both unlicensed and licensed bands. These unused spectrum bands detected through spectrum sensing show different characteristics according to the radio environment. Since CR networks can have multiple available spectrum bands having different channel characteristics, they should be capable of selecting the proper spectrum bands according to the application requirements, called spectrum decision. In this project, we propose an application-adaptive spectrum decision method over heterogeneous spectrum bands. At first, each spectrum band is characterized for the spectrum decision, based on not only local observations of CR users but also statistical information of primary networks. Through the local measurement, CR users can estimate the channel conditions such as capacity, bit error rate (BER), delay and jitter. In order to describe the dynamic nature of CR networks, we propose a new metric, primary user activity, defined as the probability of the primary user appearance during the CR user transmission. After the spectrum characterization, the CR network chooses the best spectrum bands through the following spectrum operations. the CR network uses multi-spectrum transmission based on OFDM technology. This decision process can be modeled as an optimizaiton problem.

In this project, a QoS aware spectrum decision framework is proposed to determine a set of spectrum bands by considering the application requirements as well as the dynamic nature of spectrum bands as shown in Figure 3. Specifically, for real-time applications, a minimum variance-based spectrum decision (MVSD) is proposed so as to minimize the capacity variance of the decided spectrums subject to the capacity constraint. Furthermore, a maximum capacity-based spectrum decision (MCSD) is proposed for the best effort applications where spectrum bands are decided to maximize the total throughput. Moreover, a dynamic admission control scheme is developed to decide on the spectrum bands adaptively dependent on the time-varying CR network capacity.



Figure 4. Spectrum decision framework for cognitive radio networks.< >

Related work:

W. Y. Lee, and I. F. Akyildiz, ``A Spectrum Decision Framework for Cognitive Radio Networks," Submitted for journal publication, Nov. 2007.

---

Inter-Cell Spectrum Sharing in Cognitive Radio Networks

Cognitive radio (CR) networking achieves high utilization of the scarce spectrum resources without causing any performance degradation to the licensed users. Since the spectrum availability varies over time and space, the infrastructure-based CR networks are required to have a dynamic inter-cell spectrum sharing capability. This allows fair resource allocation as well as capacity maximization and avoids the starvation problems seen in the classical spectrum sharing approaches. In this paper, a joint spectrum and power allocation framework is proposed that addresses these concerns by (i) opportunistically negotiating additional spectrum based on the licensed user activity (exclusive allocation), and (ii) having a share of reserved spectrum for each cell (common use sharing). Our algorithm accounts for the maximum cell capacity, minimizes the interference caused to neighboring cells, and protects the licensed users through a sophisticated power allocation method.




Figure 5. Inter-Cell Spectrum Sharing Framework.

Infrastructure-based CR networks are required to provide two different types of spectrum sharing schemes: intra-spectrum sharing and inter-spectrum sharing. In order to share spectrum resource efficiently, CR networks necessitate a unified framework to support cooperation among inter- and intra-cell spectrum sharing schemes and other spectrum management functions. Figure 5 shows the proposed framework for spectrum sharing in infrastructurebased CR networks, which consists of inter-cell spectrum sharing, intra-cell spectrum sharing, and event monitoring.

• Event Monitoring: The event monitoring has two different functionalities. One is to detect the PU activities, called spectrum sensing. CR users sense the radio environment continuously and send monitoring results to their base-station. Here we assume the periodic sensing which has separate time slots for sensing and transmission. In addition, CR users monitor the quality-of-service (QoS) of their transmission. According to the detected event type, the base-station determines the spectrum sharing strategies and allocates the spectrums to each user adaptively to the radio environments.
   
• Cell Spectrum Sharing: The intra-cell spectrum sharing enables the base-station to avoid the interference to the primary networks as well as to maintain the QoS of its CR users by allocating spectrum resource adaptively to the event detected inside its coverage. If a new CR user appears in this cell, the base-station determines its acceptance and selects the best available spectrum band if it is admitted. Furthermore, when some of its CR users cannot maintain the guaranteed QoS or lose their connections due to the PU activities, the base-station should re-allocate the spectrum resource to them immediately. Also a CR MAC protocol is required to allow multiple CR users to access to the same spectrum band. The intra-cell spectrum sharing has been widely investigated in many literatures and is out of the scope in this project.
   
• Inter-Cell Spectrum Sharing: In CR networks, the available spectrum bands vary over time and space which makes it difficult to provide reliable spectrum allocation. Especially in the infrastructure-based networks, the inter-cell interference also needs to be considered in spectrum sharing so as to maximize the network capacity. In the proposed framework, the inter-cell spectrum sharing is comprised of two subfunctionalities: spectrum allocation and power allocation. In the spectrum allocation, the base-station determines its spectrum bands by considering the geographical information of primary networks and current radio activities. The power allocation enables the base-station to determine the transmission power of its assigned spectrum bands so as to maximize the cell capacity without interference to the primary network. When the service quality of the cell becomes worse or is below the guaranteed level, the base-station initiates the inter-cell spectrum sharing and adjusts its spectrum allocation. Based on the spectrum allocation, the base-station determines its transmission power over the allocated spectrum bands.

Related work:

W. Y. Lee, and I. F. Akyildiz, ``Joint Spectrum and Power Allocation for Inter-Cell Spectrum Sharing in Cognitive Radio Networks," to appear in Proc. of IEEE DySPAN 2008, Chicago, IL, USA, Oct. 2008.

---

Spectrum Mobility for Cognitive Radio Networks

With their capability to support flexible usage of wireless radio spectrum, cognitive radio (CR) techniques have attracted increasing attention in recent years. In CR networks, secondary users may dynamically access underutilized spectrum without interfering with primary users, which is called spectrum handoff. Spectrum handoff refers to the procedure invoked by the cognitive radio users when they users wish to transfer their connections to an unused spectrum band. Spectrum handoff occurs 1) when primary user is detected or 2) current spectrum condition becomes worse. The cognitive radio users monitor the entire unused spectrum continuously during the transmission. If spectrum handoff occurs, they move to the "best matched" available spectrum band. However, due to the latency caused by spectrum sensing, decision and handoff procedures, quality degradation is inevitable during spectrum handoff. Hence, our spectrum handoff method focuses on the seamless transition with minimum quality degradation.

In this project, mobility management scheme for spectrum handoff in infrastructure based CR networks is proposed. A user 2D mobility profile (U2DMP) framework is defined to support a spectrum and space mobility management (S2MM). Besides space mobility characteristics and service pattern as the traditional mobility profile, spectrum mobility characteristics are included in the U2DMP special for spectrum handoff. For infrastructure based CR networks, a novel hierarchical spectrum handoff scheme with intra/inter-pool spectrum handoff is designed. The proposed adaptive QoS intra-pool spectrum handoff scheme considers both application level QoS and connection level QoS of different SUs to efficiently re-allocate spectrum resource inside a cell. The proposed adaptive threshold inter-pool spectrum handoff scheme adopts mix load indicator of PUs and SUs to get adaptive threshold to trigger network initiated handoff. Then opportunistic cell capacity is used to calculate the handoff amount, and the most suitable SUs in the overlapping areas are chosen to accomplish the handoff. As spectrum-space domain resource utilization, the combination of the two kinds of handoff can maximize spectrum utilization and minimize spectrum handoff dropping probability.

---

Spectrum Aware Routing Protocol for Cognitive Radio Networks

Routing constitutes a rather important but yet unexplored problem in CR networks, especially when a multi-hop ad-hoc architecture is considered. The activity of the primary users (PUs) affects the channels of the licensed bands differently. This renders the channels unusable for the CR network to different geographical extents around the PU. In such a situation, the key decision is switching the channel in portions of the route, thus incurring a switching delay, or passing through entirely different regions altogether, thus increasing the latency. In addition, the frequently changing primary user (PU) activity and the mobility of the users make the problem of maintaining optimal routes in ad-hoc CR networks challenging. In this project, we develop a geographic forwarding based SpEctrum Aware Routing protocol for Cognitive Ad-hoc networks (SEARCH), that:

• Jointly undertakes path and channel selection to avoid regions of PU activity during route formation.
• Adapts to the newly discovered and lost spectrum opportunity during route operation.
• Predicts node mobility and takes corrective measures to maintain end-to-end performance.

Figure 6. Joint route and spectrum discovery. We consider a three-dimensional system, with the x-y plane representing the physical space where the CR network and the PUs are located. The z-axis shows the frequency scale and also the different channel bands. The shaded regions in the figure show that a single PU may affect several channels (frequencies) around its location. Moreover, the channels may be affected to different geographical extents, depending upon their frequency separation with the PU's transmission channel. SEARCH attempts to find paths which circumvent the PU coverage regions (Path 1 and 2) and link them together, whenever a performance benefit is seen.

Route Formation:
• Paths are first constucted on each channel, independently of the others. They are then merged at the destination and these combination points indicate channel switching where a finite switching delay is incurred. SEARCH balances the path delay due to circumventing the PU region as against the switching delay in order to find the best route to the destination.



• Route Maintenance: SEARCH binds routes to regions found free of PU activity, rather than particular CR users. Thus even if nodes move away, the route is kept operational by choosing replacement nodes that are close to the original locations. Moreover, when new spectrum is detected or the channel becomes unusable owing to the appearance of a PU, the route is updated. During these update stages, the optimality of the route may be partially lost as the key consideration is prevention of performance degradation to the PU. SEARCH intelligently minimizes complete re-routing by undertaking local recovery actions.

---

Transport Protocol for Ad-hoc Cognitive Radio Networks

The dynamic nature of the underlying spectrum in cognitive radio networks is known to have adverse influences on the overall throughput of the transmission functions. It is important to seamless integrate the channel sensing and primary user (PU) detection techniques in the design of higher layer, end-to-end transport protocols. In this project, we devise a window-based transport protocol that that not only addresses the key concerns on classical wireless ad hoc networks, such as mobility, but also considers spectrum sensing, channel switching, PU activity and other issues that are commonly seen CR networks.



Figure 7. Component blocks of the transport protocol for ad-hoc CR networks.

The chief component blocks of the transport layer protocol are shown in Figure. 7. The key challenge in this research is identifying the cause of performance degradation and taking the corrective steps under changing spectrum conditions. The transport layer metrics, such as round trip time, acknowledgement (ACK) packet timeouts and the status of the congestion window need to be interpreted under CR specific concerns (channel sensing, switching, available bandwidth) and distruptions due to node mobility, apart from the case of classical network congestion. The rate control block receives feedback from the end-to-end transport layer metrics, the application layer QoS bounds and the CR network state. It then chooses the congestion window and timeout threshold parameters for the subsequent operation.

The following challenges will be considered in designing efficient transport layer protocols for cognitive radio networks. 1) The protocol must distinguish between packet delay caused by periodic channel sensing and the channel switching delay, as against congestion scenarios. This may need specific feedback from the network nodes and different algorithms to interpret the transport layer metrics under varying network conditions. 2) As the spectrum opportunity is available for a very short duration, the efficiency of the protocol needs to be high. This may necessitate scaling the transmission metrics to utilize the channel immediately to its fully capacity. 3) The wireless channel access delay in cognitive radio network depends on the frequency of operation, interference level, and medium access control protocol. Therefore, it is essential to model channel access delay in cognitive radio network for different operating scenarios and then incorporate that into the design of transport layer protocols. 4) The transport protocol needs to dynamically adapt to the changing network state. This may imply evolving the transmission threshold parameters over time. How often must these parameters change and to what extent are some of the questions that shall be answered in the course of the protocol design.

---

Cognitive Mesh Networks

The Wireless Mesh Network (WMN) paradigm is envisaged to be a key technology that allows ubiquitous connectivity to the end user. A typical WMN consists of mesh routers (MRs) forming the backbone of the network, interconnected in an ad-hoc fashion. Each MR can be considered as an access point serving a number of users or mesh clients (MCs) under it. The MCs could be mobile users, stationary workstations or laptops that exchange data over the Internet. Our proposed COgnitive Mesh NETwork (COMNET) architecture, takes the first step in leveraging the benefits of cognitive radio technology in the area of WMNs.

Figure 7. Cognitive mesh network architecture.
This project aims to address the following key challenges in a cognitive radio enabled mesh scenario:

• Enabling MCs to monitor the primary channel while continuing normal operation in the 2.4 GHz ISM band.


• Devising a theoretical framework for identifying primary transmitter frequencies through time domain sampling.


• Proposing theoretical models for estimating power injected in the primary band channels due to the presence of secondary users.


• Allowing a decentralized computation framework at each MR for load sharing between the primary and secondary bands, based on the above models.

Related work:

K. R. Chowdhury and I. F. Akyildiz, ``Cognitive Wireless Mesh Networks for Dynamic Spectrum Access ," IEEE Journal of Selected Areas in Communications (JSAC), Vol. 26, No. 1, pp. 168-181, January 2008.

---

Cognitive Sensor Networks: PU Location and Channel Detection

Wireless sensor networks (WSNs) are being increasingly deployed for a variety of environment monitoring, commercial utility metering, military and surveillance applications. These applications necessitate reliable data delivery with minimum packet loss due to external interference. When these sensors form a CR network, some of the transmission channels may be affected by the PU. The problem of sensing is more involved as the hardware restrictions on the nodes result in reduced storage and computational ability. Moreover, in a noisy environment with multiple interferer types, distinguishing between them becomes a key challenge. Localization based on triangulation with signal strength alone is also be ineffective, as their coverage regions may overlap and the effect of a single interferer cannot be isolated. At the same time, awareness of the type of the PU in a heterogeneous environment, its channel and location is beneficial as optimized routing and channel switching strategies may be constructed.

The problem of detecting the above interferer characteristics is addressed in this project by:

• Experimentally profiling the spectral characteristics specific interferer types
• Devising a scheme for identifying interferer type based on matching the observed and the previously obtained spectral data
• Estimating interferer locations by adapting the classical k-means clustering algorithm.

Figure 8. Identifying the interferer type and location.


We first conduct experiments to derive how the channels used by the sensors are affected due to specific interferers. For our experimental study, we use the WLAN and commercial microwave ovens. The measured received power in each channel is used to create a reference vector (Figure 8 (a)) in an n-dimensional space, where n is the total number of channels affected. This reference vector is then compared with an observed channel power vector during the network operation to identify the type of the interferer based on the channels that are affected. Finally, using clustering methods, the sensors that sense a common interferer are grouped together. The PU or interferer location can then be approximated as the cluster center (Figure 8 (b)). This process allows sensors to identify the PU characteristics by delegating the computational complexity to the sink node.


Related work:

K. R Chowdhury, M. Khan and I. F. Akyildiz, ``Cognitive Sensor Networks - Spectral Maps And Applications," Submitted for conference publication, June 2008.