PaperlayerView exact revision on arXiv

Source-linked AI summary

Multiband Spectrum Access: Great Promises for Future Cognitive Radio Networks

Ghaith Hattab, Mohammed Ibnkahla

arXiv:1409.5913v1cs.IT

TL;DR

Spectrum scarcity motivates cognitive radio, yet multiband operation introduces wideband-front-end and spectrum-access challenges. The paper surveys multiband sensing and access, cooperative MB-CRNs, performance metrics, and design tradeoffs, finding that cooperative multiband cognitive radio offers a compromise between spatial diversity and sampling complexity.

  • Problem

    Existing research has focused mainly on single-band cognitive radio, leaving multiband sensing, access, metrics, and associated design tradeoffs in need of unified analysis.

  • Method

    The paper surveys multiband sensing techniques, cooperative MB-CRNs, performance metrics, and fundamental limits and tradeoffs.

  • Results

    Cooperative multiband cognitive radio provides a compromise between spatial diversity and sampling complexity while multiband access can enhance throughput and channel maintenance.

  • Takeaways & Limitations

    Multiband cognitive radio offers a framework for simultaneously using multiple bands, with cooperation helping balance sensing diversity against sampling requirements.

Abstract

from arXiv · show

Cognitive radio has been widely considered as one of the prominent solutions to tackle the spectrum scarcity. While the majority of existing research has focused on single-band cognitive radio, multiband cognitive radio represents great promises towards implementing efficient cognitive networks compared to single-based networks. Multiband cognitive radio networks (MB-CRNs) are expected to significantly enhance the network's throughput and provide better channel maintenance by reducing handoff frequency. Nevertheless, the wideband front-end and the multiband spectrum access impose a number of challenges yet to overcome. This paper provides an in-depth analysis on the recent advancements in multiband spectrum sensing techniques, their limitations, and possible future directions to improve them. We study cooperative communications for MB-CRNs to tackle a fundamental limit on diversity and sampling. We also investigate several limits and tradeoffs of various design parameters for MB-CRNs. In addition, we explore the key MB-CRNs performance metrics that differ from the conventional metrics used for single-band based networks.

I. INTRODUCTION

Cognitive radio addresses spectrum scarcity through opportunistic access, but reliable sensing and coexistence with primary users remain central challenges. The paper motivates multiband cognitive radio and develops a unified survey of sensing, cooperation, metrics, and design tradeoffs.

  • I. INTRODUCTION: Cognitive radios let secondary users opportunistically access unused licensed spectrum while protecting primary-user transmissions.CRNs must manage coexistence through interweave, underlay, or overlay access paradigms.
  • I. INTRODUCTION: Spectrum sensing enables local determination of spectrum status, but sensing reliability and periodic monitoring remain necessary challenges.The paper focuses on active spectrum sensing rather than passive awareness through beacons or databases.
  • I. INTRODUCTION: Multiband cognitive radio simultaneously senses and accesses multiple channels, promising higher throughput, seamless handoff, and fewer transmission interruptions.These benefits motivate the wideband cognitive-radio paradigm.
  • I. INTRODUCTION: Deep fading can cause hidden-terminal errors, while cooperative sensing provides spatial diversity by combining observations from multiple secondary users.Cooperation is presented as a response to unreliable local sensing in shadowed conditions.
  • I. INTRODUCTION: The paper establishes a unified framework covering multiband sensing techniques, cooperative MB-CRNs, performance metrics, and design limits and tradeoffs.Its sensing discussion includes serial, parallel, wavelet, and compressive approaches.
  • I. INTRODUCTION: Spectrum sensing formulates primary-user absence or presence as binary hypothesis testing using a test statistic and threshold.The paper introduces coherent, energy, and feature detection as common single-band techniques.

2) Energy Detection:

Energy detection requires no prior knowledge of the primary-user signal, whereas feature-based alternatives exploit signal structure. Cooperative sensing combines user decisions or statistics, trading interference protection, throughput, overhead, and reliability.

  • 2) Energy Detection:: Energy detection computes received-signal energy over a time window when the secondary user lacks prior knowledge of the primary signal.Its statistic is based on the received-signal norm.
  • 2) Energy Detection:: Feature detectors exploit signal redundancy and second-order or cyclostationary statistics to improve robustness against noise uncertainty.Cyclostationarity detectors use periodic statistical properties, while eigenvalue methods exploit covariance structure.
  • 2) Energy Detection:: Detector selection depends on available prior signal knowledge: coherent detection needs full knowledge, feature detection partial knowledge, and energy detection none.The alternatives differ in processing complexity, sensing time, power consumption, and sensitivity to noise-variance estimation.
  • 2) Energy Detection:: Cooperative sensing lets secondary users share detection results to mitigate hidden-terminal errors caused by shadowing or deep fades.The figure illustrates cooperation among users as a way to improve sensing reliability.
  • 2) Energy Detection:: Hard, soft, and hybrid combining are the three main techniques for aggregating cooperative sensing information.Hard combining communicates decisions, whereas soft combining communicates original sensing statistics.
  • 2) Energy Detection:: OR, AND, and majority rules respectively emphasize minimum interference protection, higher throughput, and majority-based occupancy decisions.With K cooperating users, these rules correspond to k = 1, k = K, and k = ⌈K/2⌉.

2) Soft Combining:

Soft combining shares users’ original sensing statistics and weights them for cooperative detection, while multiband access expands opportunities but introduces sampling, sensing, and channel-selection tradeoffs.

  • 2) Soft Combining:: Soft combining shares each secondary user’s original sensing statistics without local one-bit reduction.The optimal combination uses a weighted summation of the collaborative users’ statistics.
  • 2) Soft Combining:: The optimal combination converges to equal-gain combining at high SNR and maximal-ratio combining at low SNR.The weights may use equal gain or link-SNR-based maximal-ratio combination.
  • 2) Soft Combining:: Hard combining reduces overhead but loses information through one-bit decisions, whereas increasing transmitted bits improves performance at the cost of larger overhead.A two-bit softened hard-combining scheme is described as a balance between hard and soft combining.
  • Multiband Spectrum Sensing: Single-band sensing remains a building block for multiband sensing, but practical multiband access requires additional modifications and sensing techniques.The paper frames multiband sensing as an extension rather than a direct use of single-band methods.
  • Multiband Spectrum Sensing: Multiband detection divides a wideband spectrum into multiple subbands and evaluates occupancy through separate binary hypotheses when subbands are independent.The paper presents H0,m and H1,m for each subband m.
  • Multiband Spectrum Sensing: Multiband access supports backup channels, reduces handoff frequency, and enables larger transmission bandwidth for throughput or QoS goals.These benefits motivate sensing and accessing multiple bands.
  • Multiband Spectrum Sensing: Cooperative multiband sensing distributes subband monitoring among users so the group can sense the full spectrum while reducing each user’s sampling burden.This creates a tradeoff between spatial diversity and sampling requirements.

B. Serial Spectrum Sensing Techniques

Serial sensing examines multiple bands sequentially, using tunable front ends, staged searches, or sequential tests. These methods reduce some sampling or search costs but face tuning delays, high search time, and complexity under correlated subchannels.

  • Serial sensing examines multiple bands one at a time using single-band detectors.
  • Serial sensing: Reconfigurable BPFs pass one band at a time but require wideband front ends and difficult cutoff-frequency and bandwidth control.
  • Serial sensing: Tunable local oscillators down-convert each band to a fixed intermediate frequency, significantly reducing the sampling-rate requirement.
  • Limitations: Tuning and sweeping between channels hinder fast processing, while serial sensing generally has high average search time when PU presence probability is high.
  • Other algorithms: Two-stage sensing performs coarse wideband search before fine sensing of candidate subbands, providing faster searching than one-stage algorithms when PU activity is low.
  • Parallel sensing: Parallel sensing uses multiple detectors or filter banks, while frequency-domain implementations commonly apply FFT-based energy detection with subchannel-specific thresholds.
  • Parallel sensing: Correlated subchannels exponentially increase detection complexity; a weighted energy combiner can improve reliability but assumes the correlation model is known beforehand.

D. Wavelet Sensing

Wavelet sensing detects unknown subband boundaries by analyzing spectrum singularities across scales. Its multiscale methods improve edge detection or narrowband sensitivity, but false edges, smoothing choices, and computational complexity remain unresolved issues.

  • Wavelet detectors address the impractical assumption that subband number and locations are known by detecting spectrum singularities.
  • Wavelet modulus maxima: Wavelet modulus maxima uses derivatives of the continuous wavelet transform to sharpen spectral edges and characterize them.
  • Wavelet multiscale product: Wavelet multiscale products enhance edge peaks and suppress noise, but increasing J improves reliability at the expense of additional complexity.
  • Wavelet multiscale product: Impulsive noise and spectral leakage create false edges, so thresholding can limit detected boundaries but may still leave estimation errors.
  • Wavelet multiscale sum: Wavelet multiscale sums can outperform products for narrowband signals with slowly varying power spectral density because multiplication attenuates them.
  • Open challenges: Robust wavelet sensing still requires algorithms that detect true subband edges, reject false edges, and manage smoothing-function choices at low complexity.

E. Compressive Sensing

Compressive sensing exploits spectral sparsity to sample wideband signals below the Nyquist rate. It can relax ADC and front-end requirements, but reconstruction, hardware, synchronization, and spectrum-occupancy constraints remain important limitations.

  • A 3GHz-wideband signal requires at least 6GHz sampling at the Nyquist rate, motivating sub-Nyquist compressive sensing.
  • Sparsity models: Frequency sparsity and cyclic-sparsity can support multiband sensing, including sub-Nyquist cyclostationarity detection.
  • Principle: Compressive sensing reduces sampling requirements by representing sparse signals with O ≈ L measurements and reconstructing them through sparse recovery algorithms.
  • Analog implementations: Analog-to-information converters spread analog signal content with a pseudo-random generator before low-pass filtering to reduce sampling burden.
  • Analog implementations: Multi-rate asynchronous sub-Nyquist sampling uses multiple branches with different low sampling rates to address synchronization problems across cooperating users.
  • Limitations: Compressive sensing may degrade SNR, amplify hardware nonidealities, require nonlinear reconstruction, and depend on a known sparsity basis.
  • Limitations: Compressive sensing is not effective for moderately or densely occupied spectra, limiting its applicability beyond sparse-spectrum settings.

H. Comparison

Multiband sensing techniques trade implementation speed, hardware and processing complexity, sampling requirements, and prior knowledge. Cooperative multiband sensing reduces per-user sensing load by assigning subsets of channels, while introducing diversity and coordination tradeoffs.

  • H. Comparison: Serial sensing is simple but slow for many subchannels, while parallel sensing is faster but requires more RF components and complex processing.Two-stage and SPRT-based methods can accelerate serial sensing, but may require additional components, truncation, or prior PU-signal knowledge.
  • H. Comparison: Wavelet, compressive, angle-based, and blind sensing address different needs but face false edges, sparsity assumptions, receiver requirements, or limited research.Wavelet sensing estimates subchannel boundaries; compressive sensing reduces sampling requirements; angle-based sensing uses spatial information; blind detection avoids PU prior knowledge.
  • I. Practical Implementation: Practical sensing studies expose hardware limits from noise estimation, synchronization, sampling-clock offsets, cyclic-frequency estimation, nonlinearities, power consumption, and ADC sampling rates.A multi-frame statistic was validated by implementation, while adaptive interference cancellation is needed to address wideband front-end nonlinearities.
  • IV. Cooperative Communications in Multiband Cognitive Radio Networks: Cooperative multiband sensing remains less developed than single-channel cooperation and must balance sensing coverage, spatial diversity, sampling requirements, and implementation load.Each SU can sense a subset of M subchannels; uniform diversity gives every subchannel equal coverage, whereas non-uniform diversity can prioritize important or heavily occupied bands.
  • Performance Measures: MB-CRN evaluation includes sensing performance measures such as ROC curves and network throughput, extending beyond conventional single-band metrics.The ROC plots probability of detection against probability of false alarm, while Fig. 7 illustrates uniform and non-uniform diversity.

1) Single Band:

Single-band sensing evaluates detectors through detection and false-alarm behavior, with performance depending on available signal knowledge and operating SNR. ROC comparisons cover coherent, energy, and feature detectors under different OFDM information assumptions.

  • 1) Single Band:: A false alarm occurs when the SU declares a PU present on an idle band, creating a detection-throughput tradeoff with missed detection.Higher detection probability limits PU interference, while lower false-alarm probability improves SU throughput.
  • 1) Single Band:: Detector formulas depend on whether the SU knows the transmitted signal: coherent detection uses known-signal information, whereas energy detection applies when that knowledge is unavailable.The energy-detector model treats the signal as random and uses a central chi-square distribution for the test statistic.
  • 1) Single Band:: Fig. 8 compares coherent, energy, and OFDM feature detectors under scenarios differing in knowledge of useful symbols and cyclic-prefix duration.The feature detector exploits second-order statistics of the OFDM signal.
  • 1) Single Band:: Low-SNR operation is important because PU signals may be weak at the SU receiver; the stated requirement is detection of signals as low as -114dBm.The text roughly identifies the low-SNR region as SNR ≤−10 dB.

2) Cooperative Spectrum Sensing:

Cooperative spectrum sensing combines decisions from multiple SUs, improving detection performance as cooperation increases while requiring multiband metrics that account for channel-level differences and edge-detection errors.

  • 2) Cooperative Spectrum Sensing:: Hard-combining detection and false-alarm probabilities are derived from individual SU probabilities under a k-out-of-K decision rule.Soft combining instead requires deriving the distribution of the combined statistic before calculating detection and false-alarm probabilities.
  • 2) Cooperative Spectrum Sensing:: Increasing the number of cooperating SUs improves the energy detector ROC at SNR = −10dB and N = 125, while OR logic outperforms AND logic.The comparison assumes identical individual-SU detection and false-alarm performance.
  • 2) Cooperative Spectrum Sensing:: Multiband sensing lacks a unified detection and false-alarm definition, so performance may be computed per subband or aggregated across bands.Averaging can be misleading when one channel is a poor outlier; normalized weighting can reflect channel sensitivity or importance.
  • 2) Cooperative Spectrum Sensing:: Multiband metrics can represent occupancy across channels, including probabilities involving all channels or at least i free channels.Band-occupancy degree uses actual and estimated occupancy, with ‘0’ and ‘1’ representing unoccupied and occupied subchannels.
  • 2) Cooperative Spectrum Sensing:: Edge-detection evaluation must measure boundary errors because earlier metrics assume known subchannel boundaries.Average edge-detection error uses detected-edge counts and FFT size; increasing FFT size improves detection performance.

B. Throughput Performance Measures

MB-CRN throughput measures account for imperfect sensing and transmission under interweave, underlay, or hybrid access. The modeled throughput increases with accessed channels, tighter false-alarm requirements, and sensing-based spectrum sharing.

  • B. Throughput Performance Measures: Throughput is a central MB-CRN performance measure because multiband access is intended to enhance network throughput and support QoS provisioning.The model considers imperfect sensing and allows access when the PU is absent or present under interweave and underlay paradigms.
  • B. Throughput Performance Measures: Accessing more channels increases throughput, while tighter PF A further improves throughput by reducing data interruptions.The simulation uses 6MHz subchannels, uniform power allocation, σ2 = 1, and I = −20dBW.
  • B. Throughput Performance Measures: Sensing-based spectrum sharing provides better throughput than interweave access because it permits SU coexistence with the PU.Underlay access requires power adaptation to protect PUs.

VI. FUNDAMENTAL LIMITS AND TRADEOFFS

MB-CRN design requires choosing parameters that balance competing objectives such as maximizing throughput and minimizing interference to primary users. The paper frames these choices as constrained optimization problems involving sensing, cooperation, power, and channel-related variables.

  • MB-CRN design parameters include sensing time, throughput, detection reliability, cooperating-user count, power control, and channel assignment.
  • Designers select parameter values to maximize objectives such as throughput or minimize objectives such as interference to PUs.
  • The optimization formulation uses an objective function, optimization variables, and constraint functions bounded by specified limits.
  • The framework considers M subchannels, K SUs, and N samples observed by each SU.

A. Sensing Time Optimization

Sensing duration creates a throughput–reliability tradeoff: longer sensing collects more samples and improves detection, but leaves less time for transmission. The surveyed alternatives adjust frame structure, sensing adaptivity, parallel detection, or receiver architecture to manage this tradeoff.

  • A. Sensing Time Optimization: Longer sensing duration reduces throughput by shortening the transmission slot, while collecting more samples improves detection performance.With sampling frequency f_s, the number of samples is N = τf_s.
  • 1) Different MAC Frame Structures: Increasing the number of sensing slots can reduce optimum sensing time, improve throughput, and lower false-alarm probability.
  • 1) Different MAC Frame Structures: Simultaneous sensing and transmission can lengthen both sensing and transmission periods, improve PU protection, and remove the need to optimize sensing time.Decoding errors can compromise performance, and the novel frame is superior only within a demonstrated lifetime.
  • 1) Different MAC Frame Structures: A dual-radio architecture can improve sensing and throughput but requires different channels for different SUs; otherwise quiet sensing periods remain necessary.Its drawbacks include higher cost, power consumption, and receiver complexity.
  • 2) Adaptive Sensing Time: Adaptive sensing durations reduce required samples, particularly at high SNR, thereby improving overall network throughput.One adaptive approach can miss available channels previously declared occupied.
  • 4) Sequential Probability Ratio Tests (SPRTs): Parallel SPRTs reduce samples relative to fixed-sample-size detectors, while the optimal sample count decreases as the number of channels increases.The intuitive order of sensing channels by decreasing idle probability is not always optimal under adaptive transmission.

5) Number of Cooperating Users:

Cooperation improves sensing reliability and can reduce optimum sensing time, but coordination introduces delay, bandwidth, sampling-cost, and power-allocation tradeoffs. The surveyed designs therefore optimize cooperation jointly with sensing, spectrum access, and transmit power.

  • 5) Number of Cooperating Users: Increasing the number of cooperating SUs improves detection reliability and reduces optimum sensing time, but the gain saturates as participation grows.Majority voting provides the lowest optimum sensing time and highest achievable throughput among the discussed hard-combining rules.
  • 5) Number of Cooperating Users: Collecting decisions from more cooperating SUs incurs delay, while orthogonal reporting requires additional bandwidth.Censoring and selective cooperation limit participation and can reduce overhead or power use.
  • Cooperative Multiband Tradeoffs: Increasing spatial diversity raises sampling cost, although the tradeoff becomes less severe when more SUs divide the subchannels among themselves.Sampling cost is defined as the minimum sampling-rate requirement.
  • 1) Average and Peak Transmit Powers: Under average and peak power constraints, water-filling allocates more power to higher-SNR subchannels, but peak-power allocation depends on which channels are vacant.Throughput is higher with average power because that constraint is less restrictive than peak power.
  • 1) Average and Peak Transmit Powers: Joint sensing-time and power-control schemes adapt to SU–PU distance, favoring power control at long distances and stopping transmission after short-distance detection.The reported technique outperforms adaptive power allocation at long distance and adaptive sensing time at short distance.

2) Average and Peak Interferences:

Average-interference constraints can improve secondary-user throughput while also reducing throughput losses for primary users, whereas multiband channel selection and reconfiguration expose throughput, overhead, and handoff tradeoffs. Beamforming can ease the sensing-throughput tradeoff, but requires channel-state information and antenna arrays.

  • Average and Peak Interferences: The average-interference constraint Iavg yields higher secondary-user throughput than the peak-interference constraint Ipk when Iavg = Ipk.Under Iavg, water-filling is optimal; under Ipk, truncated channel inversion is optimal.
  • Average and Peak Interferences: Iavg can also cause less throughput loss for primary users because interference diversity makes random interference levels advantageous under a convex throughput function.The comparison concerns equal average and peak bounds.
  • Beamforming: Joint beamforming reduces sensing time, improves throughput, and maintains primary-user protection, but requires CSI at both secondary-user endpoints and an antenna array.The requirement is challenging because multiple primary-user networks may not provide CSI feedback.
  • Channel Selection and Adaptation: Selecting the best available channels maximizes throughput through power and rate adaptation but requires frequent reselection and incurs high overhead.An alternative selects channels supporting the least possible transmission rate, lowering throughput while reducing reselection frequency.
  • Bandwidth and Reconfiguration: Accessing all available bands can increase throughput theoretically, but raises the probability of primary-user return and makes handoff necessary, so the number of subchannels must be optimized.Channel reconfiguration can accommodate more secondary users, but reallocation causes interruptions and delays; its gains diminish as the number of users grows.
  • Performance Tradeoffs: MB-CRN evaluation must account for tradeoffs among spectrum reliability, throughput, diversity, sampling complexity, and design parameters.The paper reviews performance measures and techniques intended to reduce the impact of these tradeoffs.
Loading 1409.5913v1…