Cognitive radio wireless sensor networks applications, challenges and research trends

Editor's Notes

• #8: power waste reduction A Game Theory Based Strategy for Reducing Energy Consumption in Cognitive WSN Energy consumption and spectrum availability are two of the most severe constraints of WSNs due to their intrinsic nature. The introduction of cognitive capabilities into these networks has arisen to face the issue of spectrum scarcity but could be used to face energy challenges too due to their new range of communication possibilities. game theory for cognitive WSNs presented strategy improves energy consumption by taking advantage of the new change-communication-channel capability. Based on game theory, the strategy decides when to change the transmission channel depending on the behavior of the rest of the network nodes. The strategy presented is lightweight but still has higher energy saving rates as compared to noncognitive networks and even to other strategies based on scheduled spectrum sensing. Simulations are presented for several scenarios that demonstrate energy saving rates of around 65% as compared to WSNs without cognitive techniques.
• #9: The system has the capability of packet loss reduction, power waste reduction, high degree of buffer management, and has better communication quality. This section discusses the advantages of using cognitive radio in WSNs. Efficient Spectrum Utilization and Spaces for New Technologies The electromagnetic spectrum is a precious gift of Nature. The amount of available useable spectrum bands cannot be increased but they can be used more efficiently. With the exception of industrial, scientific and medical (ISM) radio bands, one requires a license from the government of the respective country to utilize the radio bands. Owing to the high cost associated with spectrum licensing, many researchers and hardware manufacturers have focused on developing devices for ISM bands. Therefore, ISM bands are overcrowded limiting the development of new technologies On the other hand, many licensed spectrum bands are either underutilized or unutilized Cognitive radio wireless sensors can use the unutilized spectrum, called white spaces, without disturbing the license holders. Unlicensed users can use those bands with little or no cost, so that more technologies can be developed for these bands. Multiple Channels Utilization Most traditional WSNs use a single channel for communication In WSNs, upon the detection of an event, sensor nodes generate the traffic of packet bursts. At the same time, in densely deployed WSNs, a large number of wireless sensor nodes within the event area attempt to acquire the same channel at the same time. This increases the probability of collisions, and decreases the overall communication reliability due to packet losses, leading to excessive power consumption and packet delay. CR-WSNs access multiple channels opportunistically to alleviate this potential challenge. Energy Efficiency In WSNs, there is a large amount of power waste for packet retransmission due to packet losses. CR wireless sensors may be able to change their operating parameters to adapt to channel conditions. Therefore, energy consumption due to a packet collision and retransmission can be mitigated. Global Operability Each country has its own spectrum regulation rules. A certain band available in one country might not be available in another. Traditional wireless sensors with a preset working frequency might not work in cases where the manufactured wireless sensors are deployed in different regions. On the other hand, if nodes are equipped with cognitive radio capability, they can overcome the spectrum incompatibility problem by changing their communication frequency band. Therefore, CR wireless sensors have the potential to be operated almost anywhere in the world. Application Specific Spectrum Band Utilization Currently, the number of wireless sensors deployed for different applications has increased. In WSN, data traffic is usually correlated both temporally and spatially. When any event occurs, WSNs generate packet bursts and they remain silent when there is no event. These temporal and spatial correlations introduce to the design challenge of the communication protocols for WSN. With the intelligent communication protocols in CR-WSN, it is possible that the wireless sensors deployed for the same purpose use the spectrum of different incumbents in spatially overlapping regions. This is possible with cooperative communication among SUs, which obviously mitigates interference issues. Financial Advantages to the Incumbents by Renting or Leasing Whenever and wherever some licensed spectrum bands are not required, license holders can lease their spectrum to the SUs at low cost. This can be done while retaining the access of the incumbents on the spectrum bands whenever necessary. This is very good opportunity for those who cannot obtain a direct license for a certain spectrum due to legal or financial issues. This is a win-win approach to the incumbents and SUs. Avoiding Attacks Unlike CRWS, most off-the-shelf wireless sensors work only on particular frequency bands. Taking advantage of the wide range of spectrum usability, SUs in CR-WSNs can avoid several types of attacks.
• #11: CR-WSNs may have a wide range of application domains. Indeed, CR-WSN can be deployed anywhere in place of WSNs. Some examples of prospective areas where CR-WSNs can be deployed are as follows: facility management, machine surveillance and preventive maintenance, precision agriculture, medicine and health, logistics, object tracking, telemetries, intelligent roadside, security, actuation and maintenance of complex systems, monitoring of indoor and outdoor environments. This section discusses some of the potential areas where CR-WSNs can be deployed with examples. Military and Public Security Applications Conventional WSNs are used in many military and public security applications, such as: (a) chemical biological radiological and nuclear (CBRN) attack detection and investigation; (b) command control; (c) gather the information of battle damage evaluation; (d) battlefield surveillance; (e) intelligence assistant (f) targeting, etc. In the battlefield or in disputed regions, an adversary may send jamming signals to disturb radio communication channels In such situations, because CR-WSs can handoff frequencies over a wide range, CR-WSNs can use different frequency bands, thereby avoiding the frequency band with a jamming signal. In addition, some military applications require a large bandwidth, minimum channel access and communication delays. For such applications, CR-WSNs can be a better option. Health Care In a health care system, such as telemedicine, wearable body sensors are being used increasingly. Numerous wireless sensor nodes are placed on patients and acquire critical data for remote monitoring by health care providers. In 2011, the IEEE 802.15 Task Group 6 (BAN) approved a draft of a standard for body area network (BAN) technology. Wireless BAN-assisted health care systems have already been in practice in some remote areas of developing countries, such as in Nepal and India Wireless BAN for healthcare systems is suitable for areas, where the number of health specialists is relatively low. Medical data is critical, delay and error sensitive. Therefore, the limitation of traditional WSN, as discussed in the previous section confines the potentiality of telemedicine. The QoS may not be achieved at a satisfactory level if the operating spectrum band is crowded in convenient ‘telemedicine with BAN’. The use of ‘CR wearable body wireless sensors’ can mitigate these problems due to bandwidth, jamming and global operability, hence improve reliability. Home Appliances and Indoor Applications Many potential and emerging indoor applications require a dense WSNs environment to achieve an adequate QoS. Conventional WSNs experience significant challenges in achieving reliable communication because ISM bands in indoor areas are extremely crowded Some examples of the indoor applications of WSNs are intelligent buildings, home monitoring systems, factory automation, personal entertainment, etc. CR-WSNs can mitigate the challenges faced by conventional indoor WSNs applications. Bandwidth-Intensive Applications Multimedia applications, such as on-demand or live video streaming, audio, and still images over resource constrained WSNs, are extremely challenging because of their huge bandwidth requirements. Other WSN applications, such as WSNs in a hospital environment, vehicular WSNs, tracking, surveillance, etc., have vast spatial and temporal variations in data density correlated with the node density. These applications are bandwidth-hungry, delay intolerable and bursty in nature. Because in CR-WSN, SUs can access multiple channels whenever available and necessary, CR-WSN is very suitable for these types of bandwidth-hungry applications. Rehmani et al. reported channel bonding in CR-WSNs for such bandwidth intensive applications. Real-Time Surveillance Applications Real-time surveillance applications, such as traffic monitoring, biodiversity mapping, habitat monitoring, environmental monitoring, environmental conditions monitoring that affect crops and livestock, irrigation, underwater WSNs, vehicle tracking, inventory tracking, disaster relief operations, bridges or tunnel monitoring, require minimum channel access and communication delay. Some real-time surveillance applications are highly delay-sensitive and require high reliability. A delay due to a link failure can also occur in multihop WSNs if the channel condition is not good. On the other hand, WS nodes hop to another channel if they find another idle channel with a better condition in CR-WSNs. Channel aggregation and the use of multiple channels concurrently are possible in CR-WSNs to increase the channel bandwidth Transportation and Vehicular Networks The IEEE 1609.4 standard proposes multi-channel operations in wireless access for vehicular environments (WAVE). The WAVE system operates on the 75 MHz spectrum in the 5.9 GHz band with one control channel and six service channels. All vehicular users will have to contend for channel access and use it to transmit the information in the 5.9 GHz band. However, it still suffers from spectrum insufficiency problems. This spectrum scarcity issue and the requirements of cognitive radio in WAVE have been studied . Some preliminary works in CR-enabled vehicular communications have already been done . Vehicular wireless sensor networks are emerging as a new network paradigm for proactively gathering monitoring information in urban environments. CR-WSNs are likely to be more relevant in this field. Although this area still needs to be examined, some protocols for highway safety using CR-WSNs have been proposed . Diverse Purpose Sensing Increasingly, the use of wireless sensors in the same area for different objectives coexists. In a conventional WSN, those wireless sensors attempt to access the channel in non-cooperative manners. With the help of an efficient medium access control (MAC) protocol, CR-WSN might select different channels for different applications considering load balancing and fairness.
• #12: CR-WSNs differ from conventional WSNs in many aspects. Because protecting the right of PUs is the main concern in CR-WSN, it has many new challenges including the challenges in the conventional WSNs. 1. Detection, False Alarm, and Miss-Detection Probability The detection probability is a metric used for correct detection by CRWS regarding the absence of PUs on the channel. The miss-detection probability is a metric for CRWS failing to detect the presence of the primary signal on the channel, and the false-alarm probability is a metric for the CRWS failing to detect the absence of the primary signal. In CR-WSNs, a false alarm and miss detection can violate the right of the incumbents on the channel, which is the violation of the main principle of CRNs. The right to access a network by the incumbents should be respected in any type of CRN. A false alarm can cause spectrum under-utilization and a missed detection might cause interference with the PUs. In addition, most application areas of CRWS discussed in this paper are very critical in terms of the delay and correctness of data. In CR-WSNs, a false alarm and miss-detection causes a long waiting delay, frequent channel switching and significant degradation in throughput. The issues of the false alarm and miss-detection probability for CR ad hoc networks and IEEE 802.22 WRAN have been well studied. However, this area still needs to be explored for CR-WSNs. This area needs to be examined further to meet the research challenges of CR-WSNs. 2. Hardware CR wireless sensors have hardware constraints in terms of computational power, storage and energy. Unlike conventional wireless sensors, they have a responsibility to sense channels, analyze, decide, and act. CR wireless sensors should be capable of changing the parameter or transmitters based on an interaction with its environment , a CRWS consists of six basic units: (i) a sensing unit; (ii) a processing and storage unit; (iii) a CR unit; (iv) a transceiver unit; (v) a power unit; and (vi) a miscellaneous unit. Sensing units contain sensors and analog to digital converters (ADCs). The analog signal observed by the sensor is converted to a digital signal and sent to the processing unit. The CRWS should have cognition capability using a state-of-the art artificial intelligence technology. This capability is accommodated in the CR unit. The CR unit needs to adapt the communication parameters dynamically, such as carrier frequency, transmission power, and modulation. The unit needs to select the best available channel, share the spectrum with other users, and manage the spectrum mobility, i.e., vacate the currently using channel in the case the PU wants to use that channel. A transceiver unit is responsible for receiving and sending data. Because the energy harvesting techniques in wireless sensor nodes have developed rapidly, the energy harvesting or recharging units are optional and sensor specific. A miscellaneous unit is an application- specific additional unit, such as a location-finding unit, energy harvesting unit, and mobilizing unit, etc. Akan et al. proposed a similar hardware structure of a CR wireless sensor. Designing intelligent hardware for CR-WSNs is a very challenging issue. Many artificial intelligence techniques have been proposed to fulfill the basic principle of CR, i.e., observation, reconfiguration and cognition. Some examples include artificial neural networks (ANNs), metaheuristic algorithms, hidden Markov models (HMMs), rule-based systems, ontology-based systems (OBSs), and case-based systems (CBSs). The factors that affect the choice of AI techniques, such as responsiveness, complexity, security, robustness, and stability, are discussed in reference. Nevertheless, it is unclear how much intelligent hardware for WS-CRNs is intelligent enough and no threshold has been defined for it. Topology Changes Topology directly affects the network lifetime in WSNs. Depending on the application, CR wireless sensors may be deployed statically or dynamically. In any type of WSN, hardware failure is common due to hardware malfunctioning and energy depletion. The topologies for CR-WSNs may be the same as conventional WSNs, but they are prone to change more frequently than ad hoc CRNs. Akan et al. reported that CR-WSNs have the following topologies: (i) Ad Hoc CR-WSNs; (ii) Clustered CR-WSNs; (iii) Heterogeneous and Hierarchical CR-WSNs; and (iv) Mobile CR-WSNs. Basically, the minimum output power required to transmit a signal over a distance δ is proportional to δn, where 2 ≤ n < 4. The exponent n is closer to four for low-lying antennae and near-ground channels, as is typical in wireless sensor network communication. Therefore, routes that have more hops with shorter hop distances can be more power efficient than those with fewer hops but longer hop distances. Nevertheless, it is not always possible to find such a route in static sensor networks topology. Therefore, an adaptive self-configuration topology mechanism is important for CR-WSNs for obtaining scalability, reducing energy consumption and achieving better network performance. An adaptive self-configuration topology mechanism performs better than static topology, even though it is a challenging issue to design and implement. This area has not received much research attention. Fault Tolerance CR-WSNs should have self-forming, self-configuration and self-healing properties. In other words, whenever some nodes or links fail, an alternative path that avoids the faulty node or link must be derived. In CR-WSNs, faults can occur for a variety of reasons, such as hardware or software malfunctioning, or natural calamities, e.g., fire, floods, earthquakes, volcanic eruptions, or tsunamis etc. A CR-WSN should always be prepared to deal with such situations. There are several types of faults, such as node fault, network fault, and sink fault, etc. Souza et al. surveyed the well fault tolerance in WSNs. Hoblos et al. modeled the fault tolerance or reliability Rk(t) of a wireless sensor node using the Poisson distribution within the time interval (0,t) as follows: [Math Processing Error] where λk is the failure rate of wireless sensor node k and t is the time period. The fault tolerance is one of the challenging issues in CR-WSNs. The protocols designed for CR-WSNs should have a level of fault tolerance capability so that the overall function of the WSNs should not be interrupted. Manufacturing Costs Generally, CR-wireless sensors have been deployed in large numbers. Therefore, the cost should be significantly low. In contrast to conventional WSNs, which require less memory and computation capability, CR-WSNs require moderate memory and computational capabilities. To reduce the hardware cost, an algorithm that requires less computational power and memory should be developed. Designing such an algorithm is a challenging issue. Furthermore, CR wireless sensors should contain intelligent radio, application specific positioning systems (e.g., GPS), energy harvesting unit etc. which obviously increase the production cost. Clustering Logically grouping and organizing similar CR wireless sensors in their proximity has several advantages. Grouping sensor nodes into clusters has been pursued in WSNs to achieve the network scalability, reduce energy consumption and reduce the communication overhead. Several types of clustering exist, such as static, dynamic, single hop and multihop, homogeneous and heterogeneous. Very little work has been done in this area, which is also one of the research challenges. Channel Selection Because there is no dedicated channel to send data, sensors need to negotiate with the neighbors and select a channel for data communication in CR-WSNs. This is a very challenging issue, because there is no cooperation between the PUs and SUs. PUs may arrive on the channel any time. If the PU claims the channel, the SUs have to leave the channel immediately. Therefore, data channels should be selected intelligently considering the PU's behavior on the channel and using some AI algorithms. Scalability For some applications, CR wireless sensors should be deployed in huge numbers. Unlike conventional WS nodes, CR wireless sensors require cooperation among nodes for spectrum information sharing. This is very difficult to coordinate in a heterogeneous CR wireless sensor environment. Algorithms and protocols developed for CR-WSNs should be capable of solving these issues due to the growing size of the network. Power Consumption CR wireless sensors are power constraint devices with a limited energy source. In addition to the energy needed for spectrum sensing, channel negotiation, route discovery, transmission and reception of data packets, backoff, and data processing, CR wireless sensors also require energy for frequent spectrum handoff. A CR wireless sensor needs to sense the PUs' activities on the channel. Many applications require multiple antennas to monitor the PUs' activities, hence more energy is consumed. Although there are several proposals for energy harvesting, these techniques have their own limitations. In some application scenarios, energy harvesting or the replenishment of power resources is not possible. In ad hoc CRNs, power consumption is an important design factor, but not the primary consideration. However, in CR-WSNs, it is one of the main performance metrics that directly affect the network lifetime. Quality of Service (QoS) In conventional WSNs, the QoS is generally characterized by four parameters: bandwidth, delay, jitter and reliability. To avoid hazardous consequences in critical applications, WSNs need to maintain an adequate level of QoS. QoS support is a challenging issue due to resource constraints, such as processing power, memory, and power sources in wireless sensor nodes. This is more challenging in CR-WSNs because in addition to the challenges in WSNs, it has one more challenge to protect the rights of PUs to access the incumbent spectra. PU's communication should be interference free with the SUs. This is more challenging because it is difficult to predict the PU's arrival on the channel. A miss-detection of the primary signal and a false alarm can cause additional challenges. Security Wireless sensors are normally deployed in an unattended environment, and are prone to security and privacy issues. CR wireless sensors can be attacked physically, and the data can be stolen. CR-WSNs are more vulnerable to security threats than the conventional WSNs, because there is no strict cooperation between PUs and SUs communication. The data collected by CR wireless sensors can be sniffed, destroyed or altered by unauthorized entities. In addition, attackers can interfere with PUs transmission or prevent the use of the channel by SUs through spectrum sensing data falsification (SSDF). CR-WSNs should have some acceptable level of security robustness against these potential threats and attacks. In addition to the security issues in conventional WSNs, additional security challenges are there for CR-WSNs. Some of the security issues in CR-WSNs can be as follows: (a) unacceptable interference to licensed users; (b) prohibiting the use of idle channels for SUs; (c) prohibiting the use of common control channels by virtually creating a bottleneck problem; (d) access to private data; (e) modification of data; and (f) injection of false data. Some possible attacks in CR-WSNs are discussed below. 11.1. Physical Layer Attack A jamming attack is one type of attack on a physical layer (PHY). In this type of attack, the adversary transmits radio signals on the wireless channels to interfere with CR wireless sensors normal operations. These adversaries may have powerful and sophisticated hardware and software. If the adversary blocks the entire network, it constitutes DoS attack at PHY. Another type of attack on PHY is a tampering attack. In this, the attacker may damage the CR wireless sensors, replace the entire nodes, or part of its hardware. 11.2. MAC Layer Attack In a MAC layer attack, the attackers violate the rules defined in the protocols and disturb the normal operation, e.g., sending undesirable packets repeatedly to disturb the normal operation and exhaust battery power, using channels selfishly, and show uncooperative behavior on the priority of a cooperative MAC-layer. This may lead to DoS attacks at the MAC layer. 11.3. Routing Layer Attack In a routing layer, the attackers alter the information on routing packets and misguide the packet forwarding sensors on the networks. The same things can happen to data packets as well. The following routing layer attacks are possible in CR-WSNs (a) Wormhole attack: In a wormhole attack, an attacker builds bogus route information and tunnels the packet to another location. This creates routing loops and wastes energy. (b) Sinkhole attack: In a sinkhole attack, the adversary provides false information to the wireless sensor nodes in the networks, such as it has the shortest route or efficient route etc. (c) Sybil attack: In a Sybil attack, a single node may present multiple identities to the other nodes in a network. This may mislead the geographic routing protocols as the adversary appears to be in multiple locations at the same time. These issues are entirely open research issues, and need to be explored further to meet its research challenges. 12. Sensing Techniques One of the main objectives of imbedding a CR in a wireless sensor is to utilize the unused licensed spectrum opportunistically. Here, opportunistically means the SUs should protect the accessing right of the PUs whenever necessary. The interference of SUs to PU depends on the sensing accuracy of SUs. If SUs can sense the channels with high accuracy, interference with the PU decreases. Depending on the sensing technique, there is a tradeoff between the sensing delay and sensing accuracy. The technique that takes a long sensing time has more accuracy with the cost of delays and vice versa. Basically, there are two types of sensing techniques: (a) signal processing techniques; and (b) cooperative sensing techniques. signal processing sensing techniques for CR-WSNs can be divided further into matched filter detection, energy detection, cyclostationary feature detection and some other techniques. Similarly, cooperative sensing can be divided further into centralized spectrum sensing, decentralized spectrum sensing and hybrid spectrum-sensing techniques. 12.1. Signal Processing Techniques (a) Matched filter detection: The matched filter detection technique requires a demodulation of the PU's information signal at PHY and MAC layers, such as the modulation type and order, pulse shaping, packet format, operating frequency, bandwidth, etc. CR wireless sensors receive that information from the PU's pilots, preambles, synchronization words or spreading codes etc. The advantage of the matched filter method is that it takes a short time and requires fewer samples of the received signal. This decreases the received signal SNR and number of signal samples required. However, matched filter detection techniques consume considerable power and require high computational complexity and perfect knowledge of the target users. (b) Energy detection: Energy detection detects the signal based on the sensed energy. This does not require prior knowledge of the PU's signals. This is a very popular technique because of its simplicity. The main disadvantage of this technique is its lower accuracy. Energy detection cannot discriminate the PUs' signal from the SUs' signal. This technique cannot be used to detect spread spectrum signals and has poor performance under a low SNR. (c) Cyclostationary feature detection: In cyclostationary feature-detection techniques, modulated signals are coupled with sine wave carriers, pulse trains, repeated spreading, hopping sequences, or cyclic prefixes. The cyclostationary feature detection technique provides better performance, even in low SNR regions. This has good signal classification ability. However, it is more complex than energy detection and high speed sensing cannot be achieved. This cannot work if the target signal's characteristics are unknown. 12.2. Cooperative Sensing CR wireless sensors may encounter incorrect judgments because radio-wave propagation through the wireless channels has adverse factors, such as multi-path fading, shadowing, and building penetration. In addition, CR wireless sensors are hardware constraints and cannot sense multiple channels simultaneously. Therefore, CR wireless sensors cooperate and share their sensing information with each other to improve the sensing performance and accuracy. Three types of cooperative sensing exist: (a) centralized cooperative spectrum sensing; (b) decentralized cooperative spectrum sensing; and (c) hybrid cooperative spectrum sensing. In centralized spectrum sensing techniques, each CR wireless sensor performs its own local spectrum sensing independently and makes a decision as to whether the PU's signal is present or absent on a particular channel. The CR wireless sensors then forward their decisions to a central cooperator, such as a cluster head, collector, or server. The central cooperator fuses the received decision of the CR wireless sensors and makes a final decision to infer the absence or presence of the PU. Whenever a CR wireless sensor wants to send data, it request channel information to the central cooperator. In decentralized cooperation, the CR wireless sensors share intra-cluster information among clusters. CR wireless sensors in a cluster perform local spectrum sensing and inform the other CR wireless sensors within the clusters of their spectrum decision. There is no central cooperator; therefore it works in a decentralized manner. This scheme requires a periodic update on the spectrum information table, hence requires more storage and computation. In hybrid cooperation, CR wireless sensors share information in a decentralized manner. However, the central cooperator may request a cluster head to send channel information whenever necessary. Although, cooperative sensing has advantages over non-cooperative sensing, it requires additional computation and resources, which is a challenge in hardware constraint CR wireless sensors.