2017 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData)
Spectral Sensing Method in the Radio Cognitive Context for IoT Applications Gina Roncancio, M´onica Espinosa
Manuel R. P´erez and Luis C. Trujillo
Universidad Santo Tom´as CEA-IoT Bogot´a, Colombia
[email protected] [email protected]
Pontifica Universidad Javeriana CEA-IoT Bogot´a, Colombia
[email protected] [email protected]
Abstract—This paper describes a spectral sensing method in the radio cognitive context for IoT applications. Due to the increasing number of Internet connected devices there are big challenges in terms of scalability, adaptability, connectivity, accessibility and reliability. The necessity to generate efficient methods to access into the wireless medium is currently a big concern. Several works have addressed this issue but few works details how spectrum sensing could really help to allocate dynamically the unlicensed frequency bands for IoT applications reducing the congestion and enabling the IoT technologies. The presented work shows the results obtained of the noise floor characterization in the frequency band for an IoT service such as the case of SigFox and gives insights about how SDR systems can be applied in the cognitive context for wireless networks in IoT. Index Terms—Internet of Things (IoT), Cognitive Radio, Software-Defined Radio (SDR), Spectrum Sensing, SigFox, Cognitive Wireless Sensor Networks (CWSN).
Fig. 1. Emerging IoT application domains[3]
I. I NTRODUCTION The Internet of Things (IoT) concept has been tackled in different works [1] and has emerged due to the enormous advances in the electronics at Cyber Physical Systems (CPS) level. The CPS refer to new generation of embedded systems compatible with the Information and Communications Technologies (ICT); by permitting the interaction of the physical processes with virtual objects through wireless networks, mobile devices among others [2]. This interaction has been in the last decade very attractive for different sectors as shown in Fig. 1, which are generating a convergence of services and products supported by IoT technologies. Statistics projected by Cisco [3] shows that by 2020 the number of objects connected to the Internet will be 50 Billion (Fig. 2), while this number continuous to grow exponentially, the tendency according to various authors is such that the number of IoT devices will overcome at least twice the entire human population. This implies big challenges regarding scalability, adaptability, connectivity, accessibility and reliability of the physical devices employed in the wireless networks. Since the number of sensor nodes deployed in studying a physical phenomenon can be in the hundreds, thousands or millions depending on the application and it is worth to notice how this possible millions of connected devices will have access to the wireless medium [4]. 978-1-5386-3066-2/17 $31.00 © 2017 IEEE DOI 10.1109/iThings-GreenCom-CPSCom-SmartData.2017.116
A large percentage of radio links for IoT devices make use of ISM bands (Industrial, Scientific and Medical) which provides an unlicensed communication in most countries. Technologies such as SigFox have generated free band solutions at the IoT user level. One of the greatest challenges to reach scalability of the connected objects in a wireless network is to characterize the frequency bands to generate future solutions without signal interference reaching convergence of different technologies and applications. Given that context, this paper presents the work in progress of a spectral detection method for characterizing the noise floor within the ISM band by means of a cyclostationary technique. The sensing technique is currently developed in the cognitive radio context based on a Software Defined Radio (SDR). The results obtained in this work were developed prior to the implementation of the network at Centro de Excelencia de Internet de las Cosas (CEA-IoT) at Bogot´a D.C. Colombia. This paper is distributed as following: Section II gives a short review on SigFox technology; section III introduce the concept of cognitive radio and gives some details of SDR, the definition of spectral detection, spectral analysis and spectral decision; section IV describes the employed spectral detection model; section V exposes the details of the proposed sensing spectrum method for the ISM in the SigFox band specifically and shows the results of the noise floor characterization; finally 756
Fig. 4. Spectrum band distribution for SigFox Technology in Europe [7]
In SigFox, transmitting messages from sensor nodes to the Base Station (BS) does not imply the use of a specific channel within this band, on the contrary, band allocation is dynamic so sensors can send information through any of the available bands. Additionally, to ensure the arrival of the message the sensor send the data in three different bands providing a reduction of the fading effects [7]. SigFox uses RFDMA (frequency division multiple random access) that solves problems of high sensitivity in the system. In addition, there is no a spectrum sensing stage before transmitting the massage, thereby saving energy used but having greater interference. The energy consumption required for network is about 50 microwatts, however the speed of sending data is limited to 100 bps and the maximum message size is 10 Byte, the range of energy consumption varies between 20 mA and 70 mA when it is in operation, and zero energy consumption when it is in inactive period [6], [8].
Fig. 2. IoT connected devices forecasting [3][3]
section VI shows the main conclusions of this work. II. S IG F OX SigFox is a communication system that uses ultra-narrow band (UNB) for IoT devices, this new technology is based on cellular networks, but its variation consists than instead of offering services with high bandwidth requirements, it allows sending messages intended for IoT that require bandwidths smaller than 1 kHz [5]. This network is deployed through a star cell infrastructure and it is completely independent of networks that currently exist, providing advantages such as long-range, ubiquity, lower power consumption and lower cost [6], [7]. Additionally, SigFox has a cloud platform that allows viewing the transmitted messages for IoT services as shown in Fig. 3.
III. C OGNITIVE R ADIO Cognitive radio is defined as a technology that allows modification of the parameters of transmission and reception in the transmitters and receiver radios through the interaction with the radio spectrum. The hardware architecture of a cognitive radio is based on a software-defined radio (SDR) which is a technology with a high digital processing and a radio frequency (RF) front-end capability for generating different transmission or reception protocols. SDR have become very popular for military and commercial applications [9], [10]. Regarding to a more detailed architecture of the SDR, Joseph Mitola [11] describes the SDR architecture as an interface between the RF and the digital signals by means of an analog to digital converter (ADC). This architecture provides to the SDR a cognitive capacity that consists in the ability of capturing information within their environment to create the best opportunities for dynamic access to the radio spectrum. This process is developed through a continuous monitoring of power density of the band within the sensed spectrum, setting the band portions that are not in use in certain geographical areas, both in time and frequency. The SDR reconfiguration capability allows to test different methods for dynamically transmitting and receiving in different bands of the radio spectrum in time and phase [12].
Fig. 3. Overview of SigFox network architecture
This technology is executed on unlicensed frequency bands, 868 MHz for Europe and 902 MHz for EE.UU according to the regulations of each country where the service will be established. Each sensor signal requires a bandwidth of 100 Hz, thus the spectrum band used for SigFox is divided into 400 channels of 100 Hz each one, the Fig. 4 shows the distribution of the SigFox channels over the spectrum used for Europe, where channel 0 corresponds to 868.180 MHz and channel 400 correspond to 868.220 MHz.
757
Cognitive radio aims to use the unused spectral spaces called spectral holes [13]. In order to develop cognitive capacity according to the steps shown in the Fig. 5 [14]. There are three basic processes that a cognitive radio performs described as follows. 1) Spectral Detection: Where a cognitive radio monitors the bands that are enabled and catch information in order to find spectral gaps. 2) Spectral Analysis: Where a cognitive radio performs a characterization of the spectral holes that are monitored in the spectral detection. 3) Spectral Decision: Where a cognitive radio defines the transmission rate and bandwidth in order to use the spectrum according with the characterization of the spectral detection and its analysis.
The cyclostationary is a method uses for sensing primary user through the explotation of the cyclics characteristics of the signal. The cyclostationary characteristics are caused by the periodicity in the signal. This sensing is used for signals en the spectrum and is abled to identifer the AWGN noise of primary user. Accoding to [17], x(n) is cyclostationary with mean μx := E{x(n)} and covariance Cxx (n; τ ) := E{[x(n) − μx (n)][x(n + τ ) − μx (n + τ )]}. For the values C¯xx (n; τ ) := Cxx∗ (n; τ ) where ∗ denotes complex conjugation, n and τ are in the set of integers Z. The signal x(n) is defined cyclostationary iff there exists an integer P such that Cxx (n; τ ) = Cxx (n + lP ; τ ) ∀n, l ∈ Z and the smallest of all the values the P is called period. To consider the period signal is performed the expantion of harmonic components whose cyclic are defined: Acxx := {αk = 2πk P , k = 0, ..., P − 1} and its Fourier coefficients cyclic correlations are related by: P −1 2πknj FS P −−→ Cxx ( 2πk Cxx (n; τ ) = k=0 Cxx ( 2πk P ; τ )e P ; τ) = 2πknj P −1 1 P . n=0 Cxx (n; τ )e P A process almost cyclostationary [17] is defined iff its mean and correlation are almost periodic sequences. For x(n) with mean cero and real, he time-varying and cyclic correlations are defined as the generalized Fourier Series pair: FS cxx (n; τ ) = αk ∈Ac Cxx (αk ; τ )ejαk n −−→ Cxx (αk ; τ ) = xx N −1 limN →∞ N1 n=0 cxx (n; τ )e−jαk n . V. P ROPOSED SENSING SPECTRUM METHOD The set up for the proposed spectral detection method is shown in the diagram of Fig. 6, it consists of an SDR platform where the frequencies, samples and screen captures of the radio spectrum portions are set, the captured information is then processed by a spectral detection system that uses spectral correlation algorithms showing the spectrum components and the radio noise floor within the SigFox frequency band.
Fig. 5. Cognitive capacity scheme for Radio [14]
The spectral detection is highlighted into the processes of cognitive capacity, the employed spectral detection model is presented in the next section. IV. S PECTRAL D ETECTION M ODEL A model for spectral detection is presented in this section for characterization of the frequency band between 902.280MHz and 902.320MHz. The band was chosen in order to characterize the Additive White Gaussian Noise (AWGN) for the SigFox service at a fixed point. The capture of a composed matrix of spectral components in the band is defined as ⎛ ⎞ P1 (t1 , f1 ) P1 (t1 , f2 ) · · · P1 (t1 , fi ) ⎜ P2 (t2 , f1 ) P1 (t2 , f2 ) · · · P1 (t2 , fi ) ⎟ ⎜ ⎟ y=⎜ ⎟ (1) .. .. .. .. ⎝ ⎠ . . . . Pn (tn , f1 )
Pn (tn , f2 )
···
Pn (tn , fi )
Where tn is the capture time, fi is the capture frequency variation and Pn represent one single capture. The AWGN whose probability density function is defined in ( 2 ), where μ and σ are the mean and variance of the Gaussian noise distribution. P DFw (x) =
1 −(x−μ)2 /2σ2 e 2πσ
Fig. 6. Spectral detection method set-up
The most relevant characteristics of the equipment used for the development of the system described in the Table I. For this research was used a daughter card of 50 MHz to 2.2 GHz compatible with the SDR. In addition, Table II shows
(2)
758
TABLE I E QUIMENTS C HARACTERISTIC Equiment
Characteristic Frecuency 0 to 6GHz. Linux, Windows and MAC compatible. MIMO. Spartan 3A-DSP 3400 FPGA.
USRP N210
the description of the SDR data reception software and digital signal processing systems. TABLE II S OFTWARE COMPONENTS Software Simulation Software
Library MATLAB
Description MATLAB 2014 DSP System ToolBox. Communications System ToolBox,USRP(R) Radio Signal Processing Toolbox Communications System ToolBox.
Fig. 8. Sigfox Full band spectral behavior
Thus, it was applied the correlation model shown in Fig. 9; which consists of comparing the signal obtained from SDR with a AWGN, the obtained result from the model is an correlation coefficient that indicate the level of similarity with the signal of interest. This value is between 1 and -1, where 1 indicates a high level correlation in-phase and -1 a high level of correlation out of phase. Performing the corresponding tests, a statistical analysis was made, where the average, standard deviation and variance of the values obtained from the correlation coefficients of some of the screen captures taken by steps frequency of 20 Hz and 100 Hz with regard to AWGN were evident, noticing a high level of correlation as shown in Table III and Table IV, respectively.
In order to test the implemented method, first some preliminary spectrum measurements were performed by means of a spectrum analyzer, where spectral behavior in the band SigFox at a fixed point of CEA-IoT was evidenced, as shown in Fig. 7 and Fig. 8
TABLE III M APPING OF CORRELATION GAUSSIAN NOISE WITH 20 H Z SCREEN CAPTURES
C1 vs Gaussian Noise
C2 vs Gaussian Noise
Average (μ)
0.9587
0.9588
Standard Deviation
8.6842e-04
8.2977e-04
Variance (σ 2 )
7.5415e-07
6.8852e-07
TABLE IV M APPING OF CORRELATION GAUSSIAN NOISE WITH 100 H Z SCREEN
Fig. 7. Spectral behavior in the SigFox band
CAPTURES
From the spectrum analysis performed in the SigFox Band with its center frequency set to 902.3 MHz and a bandwidth of 40 kHz, it was found that there are no carriers; in addition the noise floor is beyond -97 dB. On the other hand, a model was developed applying cyclostationary through MATLAB 2014 and a set of functions from the libraries of the Telecommunications toolbox in order to analyze the SigFox band set in the 902.3 MHz frequency where it was captured from 902.280 MHz to 902.320 MHz with 100 Hz frequency steps.
C1 vs Gaussian Noise
C2 vs Gaussian Noise
C3 vs Gaussian Noise
Average (μ)
0.9679
0.9678
0.9683
Standard Deviation
0.0014
0.0016
0.0012
Variance (σ 2 )
1.9180e-06
2.5388e-06
1.4596e-06
Finally the time variation model for cyclostationary signals was applied, where the first capture compared to the following
759
Fig. 9. Correlation Model
VI. C ONCLUSION
captures and spectral coefficients obtained by statistical analysis (average, standard deviation and variance) was analyzed such as shown in Table V and Table VI, noticing high level of correlation as well.
C1 vs C2
C1 vs C3
C1 vs C4
A work in progress of a spectral sensing method in the radio cognitive context for IoT applications was presented in this paper. The work addressed the necessity of sensing the spectrum for characterizing unlicensed bands where several IoT services will be deployed in the next years. In particular, this paper is oriented to the featuring of one of the ISM bands where emerging technologies like SigFox are planning to operate. Noise floor measurements with the proposed method shows excellent agreement compared with the ones obtained with a spectrum analyzer, in the band of 920 MHz. Future works points to the develop of Cognitive Wireless Sensor Networks (CWSN) techniques for energy consumption reduction, selforganization and auto-configuration. The CEA IoT will work in cognitive radio as an opportunity in the recognition of spectrum environment in IoT devices to improves networks perfomance Low power Wireless Personal Area Networks (LowPAN) and White Space TV (TVWS) in different applications like smart cities, medical, logistic and smart grid.
Average (μ)
0.9794
0.9796
0.9798
VII. ACKNOWLEDGMENT
Standard Deviation
9.6529e-04
9.0672e-04
8.8766e-04
Variance (σ 2 )
9.3178e-07
8.2214e-07
7.8793e-07
TABLE V C YCLOSTATIONARY SPECTRAL DENSITY WITH FREQUENCY STEPS OF 20 HZ C1 vs C2 Average (μ)
0.9764
Standard Deviation
4.8022e-04
Variance (σ 2 )
2.3061e-07
TABLE VI C YCLOSTATIONARY SPECTRAL DENSITY WITH FREQUENCY STEPS OF 100 HZ
The authors would like acknowledge the cooperation of all partners within the Centro de Excelencia y Apropiacin en Internet de las Cosas (CEA-IoT) project. The authors would also like to thank all the institutions that supported this work: the
760
Colombian Ministry for the Information and Communications Technology Ministerio de Tecnologas de la Informacin y las Comunicaciones - MinTIC and the Colombian Administrative Department of Science, Technology and Innovation Departamento Administrativo de Ciencia, Tecnologa e Innovacin Colciencias through the Fondo Nacional de Financiamiento para la Ciencia, la Tecnologa y la Innovacinn Francisco Jos de Caldas (Project ID: FP44842-502-2015). R EFERENCES [1] A. Luigi, I. Antonio and M. Giacomo, The Internet of Things: A Survey, in: Computer Networks, vol. 54, nm. 15, pp. 27872805, 2010. [2] E. Borgia, The Internet of Things Vision: Key Features, Applications and Open Issues, in: Computer Communications, vol. 54, pp. 1-31, 2014. [3] D. Evans, The Internet of Things, How the Next Evolution of the Internet is Changing Everything, Whitepaper, Cisco Internet Business Solutions Group, 2011. [4] S. Cho and A. P. Chandrakasan, Energy Efficient Protocols for Low Duty Cycle Wireless Microsensor Networks, in: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), vol. 4, pp. 20412044, 2001. [5] K. E. Nolan, W. Guibene and M. Y. Kelly, An Evaluation of Low Power Wide Area Network Technologies for the Internet of Things, in: Wireless Communications and Mobile Computing Conference (IWCMC), pp. 439444, 2016. [6] A. Ali, G. A. Shah and J. Arshad, Energy Efficient Techniques for M2M Communication: A Survey, in: Journal of Network and Computer Applications, vol. 68, pp. 4255, 2016. [7] G. Margelis, R. Piechocki, D. Kaleshi and P. Thomas, Low Throughput Networks for the IoT: Lessons Learned from Industrial Implementations, in: IEEE 2nd World Forum on Internet of Things (WF-IoT), pp. 181186, 2015. [8] Z. D.R. Gnimpieba, A. Nait-Sidi-Moh, D. Durand and J. Fortin, Using Internet of Things Technologies for a Collaborative Supply Chain: Application to Tracking of Pallets and Containers, in: Procedia Computer Science,vol. 56, pp. 550557, 2015. [9] R. Akhtyamov et al., An implementation of Software Defined Radios for Federated Aerospace Networks: Informing Satellite Implementations using an Inter-balloon Communications Experiment, in: Acta Astronautica, vol. 123, pp. 470478, 2016. [10] J. Bonior, Z. Hu, T. N. Guo, R. C. Qiu, J. P. Browning and M. C. Wicks, Software-Defined-Radio-Based Wireless Tomography: Experimental Demonstration and Verification, in: IEEE Geoscience Remote Sensing Letters, vol. 12, nm. 1, pp. 175179, 2015. [11] J. Mitola, Software Radio Architecture: a Mathematical Perspective, in: IEEE Journal on Selected Areas in Communications, vol. 17, nm. 4, pp. 514538, 1999. [12] I. F. Akyildiz, B. F. Lo and R. Balakrishnan, Cooperative Spectrum Sensing in Cognitive Radio Networks: A Survey, in: Computer Physics Communications, vol. 4, nm. 1, pp. 4062, 2011. [13] S. Haykin, Cognitive Radio: Brain-Empowered Wireless Communications, in: IEEE Journal on Selected Areas in Communications, vol. 23, nm. 2, pp. 201220, feb. 2005. [14] I. F. Akyildiz, W. Y. Lee, M. C. Vuran and S. Mohanty, NeXt generation/dynamic Spectrum access/cognitive Radio Wireless Networks: a Survey, in: Computer Networks, vol. 50, nm. 13, pp. 21272159, 2006. [15] W. A. Gardner, Exploitation of Spectral Redundancy in Cyclostationary Signals, in: IEEE Transactions on Signal Processing Magazine, vol. 8, nm. 2, pp. 1436, 1991. [16] D. M. Martinez and A. G. Andrade, Adaptive Energy Detector for Spectrum Sensing in Cognitive Radio Networks, in: Computers and Electrical Engineering, 2015. [17] Madisetti, Vijay K and Williams, Douglas B and others,The digital signal processing handbook,1998.
761