Software Tools: LabVIEW NXG and MATLAB
- H. Joshi, M. Alaee-Kerahroodi, A. A. Kumar, B. Shankar Mysore R and S. J. Darak, “Learning Based Reconfigurable Sub-nyquist Sampling Framework for Ultra-wideband Angular Sensing,” ICASSP 2020 – 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2020, pp. 4637-4641. Click here to download presentation slides!
- H . Joshi, S. J. Darak and A. A. Kumar, “Low-Complexity Reconfigurable and Intelligent Ultrawideband Angular Sensing,” in IEEE Systems Journal, Jan. 2020. (Early Access)
- H. Joshi, S. J. Darak, A. A. Kumar and R. Kumar, “Throughput Optimized Non-Contiguous Wideband Spectrum Sensing via Online Learning and Sub-Nyquist Sampling” in IEEE Wireless Communication Letter, vol. 8, no. 3, pp. 805-808, June 2019.
In this work, an intelligent and reconfigurable ultra-wideband angular sensing (UWAS) framework is proposed which unlike the existing UWAS methods is independent of the maximum number of active transmissions in a wideband spectrum. To perform this, we propose a sub-Nyquist sampling and sparse ruler based multi-antenna array receiver architecture. By characterizing and selecting a set of frequency bands via learning algorithm, the proposed receiver allows sensing of more active transmissions than the number of antenna over an unlimited bandwidth. The simulation results show that due to the learning based approach, the proposed UWAS outperforms when compared to non-learning based UWAS method.
Based on the various spectrum measurement and surge of multi-antenna as well as beamforming techniques, various sub-Nyquist sampling (SNS) based ultra-wideband angular sensing (UWAS) to estimate spectrum utilization and direction-of-arrival (DoA) have been studied. However, existing UWAS methods suffer from two limitations:
- Their architecture requires the prior knowledge of the maximum number of active transmissions in the wideband spectrum which is difficult to predict in real-networks, and
- They perform contiguous UWAS which is computationally expensive and inefficient.
To overcome these limitations we used the online learning based framework, which characterizes the spectrum occupancy to optimize the throughput (i.e. the number of identified transmission opportunities) by digitizing the frequency bands that are more likely to be vacant. Since the sampling rate of SNS increases with spectrum bandwidth, we propose a sparse ruler and finite rate of innovation (FRI) modelling based multi-antenna receiver architecture. By maintaining the sampling rate, FRI modelling digitize a set of non-contiguous frequency bands over an unlimited bandwidth whereas a sparse ruler creates virtual antennas which allow sensing of more number of active transmissions than the number of antennas. The integration of learning algorithm with such architecture offers a significant improvement in performance as compared to state-of-the-art approaches.
To validate the performance of the proposed receiver architecture, we develop a hardware testbed to determine the DoAs of target signals present in the sensed multi-band signal. Here, we develop the receiver architecture for uniform linear array (ULA) and sparse antenna array arrangement. We vary the number of target signals from one to three and correspondingly vary the number of receive antenna from two to four.
Fig. 1 shows the block diagram of the proposed hardware testbed and it consists of:
- Target Transmitter: Uses two NI-USRP 2944R for the transmission of three targets signals.
- Receiver sub-System: Uses two NI-USRP 2944R to support 4-antenna Sparse/ULA antenna arrangement.
- Synchronization sub-System:
- Time and Frequency Synchronization: Uses one Ettus Octoclock for generating common reference clock and PPS signal to receiver sub-system.
- Phase Synchronization: Uses one Ettus USRP N200 for the reference signal transmitter.
- Host System: For the baseband signal processing of transmitter, receiver and reference USRPs.
- Carrier Frequency: 2GHz
- IQ Sampling Rate: 16MHz
- Tx Antenna Gain: 0
- Rx Antenna Gain: 0/2/4/6/10 dB
- A sinusoidal wave of carrier frequency 10Hz is used.
- Multiband signal consisting of SC-FDMA signal is transmitted
- A multi-band signal consisting of 8 frequency bands is considered. The busy status of these bands vary with user-defined occupancy statistics and is unknown to the receiver.
- When a band is busy, an SC-FDMA signal is transmitted.
- Works in two stages:
- Phase calibration due to individual channel noise
- DOA Estimation
- Receives two inputs:
- A reference signal (at 10 kHz) via SMA cables for phase calibration
- Target multiband signal via designed antenna array for DOA estimation.
Fig. 2 shows the frequency spectrum of the reference and multiband target signal received at the receiver
- Inputs are combined at the receiver port of the USRP RIO with the help of combiner
- Perform phase calibration
- To extract the reference signal, the received reference + target signal is low pass filtered
- The phase offset with respect to the first receiver channel of USRP is calculated for other three USRP receiver channels
- The calculated phase offset is subtracted from the phase of the received and filtered multiband target signal at the other three antennas.
Joint Estimation of carrier frequency and DOA is performed.
Fig. 3 shows the hardware set-up where we are transmitting directional multi-band traffic. At the receiver, we are first detecting vacant and busy frequency bands followed by the determination of carrier frequency and DOA of the detected busy bands. In the figure we are detecting one busy bands and the MUSIC Spectrum is showing the corresponding DOA estimated over Nyquist and sub-Nyquist samples.
Fig. 4 shows the LabVIEW front panel output when the multi-band signal transmits one target signal, two target signals and three target signals in a multi-band signal. The results show the direction of arrival (DOA) of the detected target signals. Please note that DOA_NS and DOA_SNS denote the DOA estimated via Nyquist samples and Sub-Nyquist samples respectively.