You are here: Project > Technical Highlights > Radio Environment Mapping and Sensing > 
27.3.2017 : 12:35

WP3 advances against its set objectives and summary of technical highlights is as follows:

Modelling of Radio Environment

The aim of this task is to perform measurements about spectral activity, model the activity and finally emulate the radio scene. This has led to the development of the original integrated so-called MME framework, standing for Measurements, Modeling and Emulation. In its simple form, this integrated approach consists in emulating models derived from measurements in order to stimulate the cognitive radio system under test with a realistic radio scene, similar to what it would experience in the field. The framework has been developed and proposed in a paper submitted to IEEE Vehicular Technology Magazine [1].

 

 

Figure1: MME approach

 

Such integrated approach guarantees to the user a high-fi mimicking of the radio scene which is crucial for reliable system assessment. Different flows targeting different goals can be envisaged. Several measurement campaigns have been performed, over different bands and locations, brief summary is as follows:

  • ISM band measurements (power based) by OULU University
  • Wideband measurements (power based) with main focus on GSM by Aveiro University
  • GSM and UMTS measurements (with decoding) by Thales in Paris
  • TV White Space measurements by UniS in Guildford, UK

 

 

 

Figure2: Exemplary measurement results [GSM band, Aveiro]

 

These campaigns has allowed to developed models for spectrum occupancy and more accurately traffic occupancy along the day for various standard. They have been also used to validate on real environment some of the spectrum sensing algorithm developed such as Localization Algorithm based on Double-thresholding or Sensing in Code.

In additional, Agilent developed a “radio scene emulator” which aims at emulating a given radio scene. This tool (for which a patent application was filed) gives full freedom to the user along the time/frequency/waveform/power axis/domains.

 

Review of Protocol Stack Messages for Radio Environment

 Within this task we established a framework for implementing sensing functionalities and a protocol stack for exchanging sensing information.

 Based on QoSMOS system requirements, LTE extension in TVWS and extension to 802.11, specific spectrum sensing requirements were established. They concern the measurement phase (How and when sensing is performed and controlled), the reporting phase (exchange of sensing results) and performance. These requirements are the baseline to establish the spectrum sensing architecture and interfaces, the message sequence charts and the protocol stack messages.

 Based on the system reference model of QoSMOS as defined in WP2 and on the spectrum sensing requirements, a specific model for the spectrum sensing was extracted. It incorporates the basic functionalities and builds a framework for all the activities related to sensing within QoSMOS. The model also highlights the different interfaces needed to communicate with other system blocks and with other sensing blocks. It is shown in Figure3. It consists of two main blocks, the SPECTRUM SENSING SENSOR CONTROL (SSS-CTRL) and the SPECTRUM SENSING MANAGEMENT (SS-MGT) block. Within these two blocks other blocks are incorporated which perform the sensing measurements and decisions and which use data-bases for having access to different models or algorithms. Also the interfaces to external entities are defined in the model.

 

 

Figure3: Reference model for sensing

 Spectrum Sensing Message Sequence Charts were detailed for various types of sensing configurations that are:

  • local spectrum sensing,
  • centralized spectrum sensing (the decision is taken in the fusion centre),
  • distributed spectrum sensing (Any node is able to generate a sensing decision (a SS SCTRL and SS MGT in each node)
  • interference monitoring (Interference level and incumbent signal level measurement for incumbent protection)

 Then, detailed spectrum sensing protocol stack messages were defined.

 

Algorithms for Radio Context Acquisition

 A complete set of algorithms to acquire and process radio environment information has been provided. This ranges from classical radio context acquisition (i.e. local sensing) to disseminated sensing with exchange of collaboratively or cooperatively collected context data.

A plethora of local sensing algorithms is proposed in the literature. They are classically sorted in three main categories: energy detection, matched filtering and cyclostationary feature detection. New approaches, based on these three categories are investigated in this deliverable as well as their context of usage:

    • Statistical test theory: For energy detection, in order to improve noise level estimation when it follows some distribution properties, statistical test theory is useful. Several tests are proposed including Anderson-Darling and Kolmogorov-Smirnov approaches. It has been shown that the detection probability is almost the same as for a perfect energy detection - which is not affected by any noise uncertainty.
    • Improved energy detection: Based on speech characteristics, the energy detection algorithm for PMSE can be improved.
    • Generalized higher-order cyclostationary feature detector: For PMSE, the 4th order cyclostationarity detector outperforms the 2nd order cyclostationarity detector when using shaping functions.
    • Hybrid detection: It consists in mixing energy detection and cyclostationary feature detection. It has been developed for both PMSE and OFDM. It will be proved to outperform the performance of both techniques taken separately.
    • Parallel sensing based on antenna processing: Classical cyclostationary feature detection is extended to the multi-antenna case. It will be shown that smart antenna techniques not only improve the detection sensitivity but also allow non-OFDM spatial rejection permitting an opportunistic user to perform sensing on incumbent network without managing quiet periods.
    • Sensing on GFDM: GFDM signals have been presented in WP4. As a joint contribution with WP4, WP3 studied the cyclostationary properties of GFDM signals. Improved detection probability on peaks specific to cyclostationary GFDM characteristics has been shown. As opposed to OFDM, GFDM also allows lowering the cyclic prefix and thus increasing the total amount of transmitted data, without affecting the detection probability.
    • Sensing in code: For CDMA standards, the downlink bands are always fully used by common channels even if no user is using them. This means that there might be a potential available resource (unused orthogonal codes) that classical sensing algorithms cannot detect. In order to use these codes, the opportunistic system must be able to detect them (that is to detect the used spreading factors and spreading codes). Several algorithms for the detection of UMTS and HSDPA downlink traffic channels are proposed and their performance evaluated.
    • Opportunistic system detection by watermarking: The detection of opportunistic users may be considered by the joint approach of the design of the physical layer of opportunistic users and the design of their detection. The proposed solution relies on the following proposal: the detection can be considered by the explicit introduction of specific signatures (or watermark) in the transmitted signal.
    • Blind frequency localisation: The proposed method is the Localization Algorithm based on Double-thresholding (LAD) with Adjacent Cluster Combining (LAD ACC). The iteratively operated LAD method is able to operate in any transform domain, for example, in the frequency domain. The only requirement is that in the considered data there have to be some noise samples. That is, the signal is not allowed to cover the whole studied band.

In the case of the TV white space reuse, classical sensing (meaning the detection of the presence of a signal in a given frequency band) might not be sufficient. Indeed, to verify that the reuse of the TV channel will not cause harmful interference to the incumbent receiver, the opportunistic system has to estimate the Carrier to Interference Ratio (CIR) of the incumbent receiver to determine the allowable transmit power. In order to do so a reliable CIR estimation technique is presented.

Distributed sensing is often presented as an alternative to improve local sensing performance in presence, for example, of shadowing (presence of an obstacle between the sensing device and the emitter to be sensed). The issues addressed here are no longer the detection performance by itself but how to transmit the sensing metrics to the fusion centre and how to merge numerous / various sensing metrics in an optimal way. In this deliverable, several key aspects of distributed sensing are addressed: 

    • The quantization of the metrics sent to the fusion centre: What is the metrics resolution to use in order not to degrade sensing performance and not to use too much resource for metrics transmission.
    • The impact of the time variance caused by the mobility (key point of the QoSMOS project) of cognitive radios on the sensitivity of spectrum sensing algorithms. It is a new approach as most of the actual works suppose a stationary cognitive radio. Mobility driven collaborative spectrum sensing mechanism using Neyman-Pearson’s criterion is considered and a framework for local spectrum sensing is proposed in order to exploit spatio-temporal diversity due to the user mobility.
    • How can the fusion centre deal with malfunctioning or misbehaving sensing nodes? A solution based on a beta reputation system is proposed. The performance of the proposed credibility based scheme is compared with the case of equal weight combining.

 

References

 [1] "MME approach for CR Systems Evaluation: Measurement, Modelling and Evaluation in Cognitive Radio Network”, Deepaknath Tandur, Jonathan Duplicy, Kamran Arshad, Klaus Moessner, David Depierre, Janne Lehtomäki, Keith Briggs, Luis Gonçalves and Atilio Gameiro, published in IEEE Vehicular Technology Magazine;