In ultra-wideband (UWB), angle information is acquired by measuring the phase difference of arrival (PDOA) at two adjacent antennas. Unfortunately, this acquisition is quite fragile due to phase errors and works only in a limited angular range. To address the above limitations, this paper proposes a convolutional neural network (CNN)-Transformer-PDOA estimator capable of estimating direction impinging across the range of -90° to +90°. The proposed architecture first employs a lightweight CNN to automatically extract local spatial features directly from raw channel impulse response (CIR) signals without human intervention. The outputs are then fed into two layers Transformer encoder. Because its unique self-attention mechanism and positional encoding, the model can explicitly encode the temporal order of time steps and capture dynamic changes of CIR signals in the time dimension. Finally, an adaptive feature fusion module dynamically combines the CIR embeddings with PDOA measurements by learnable weight parameters. Experimental evaluations demonstrate that the CNN-Transformer-PDOA architecture reaches a mean error of only 2.9°. Compared with classical PDOA methods, it expands the angular measurement range to 180° while reducing errors by 19.03°. Compared to CNN based approaches, the error is decreased by 5.82°. Furthermore, compared to implementations without PDOA observations, this method reduces errors by 4.15°.

In complex indoor environments, non-line-of-sight (NLOS) propagation severely degrades the precision of ultra-wideband (UWB) ranging. Existing NLOS identification methods primarily rely on statistical features or waveform analysis from single-modal channel impulse response (CIR) data, but could be failed when the NLOS CIR is similar as the LOS one. Fortunately, image vision can provide more abundant information about the ranging environment and offer spatial features that can be coordinated with CIR temporal characteristics. Based on the above insight, this paper proposes a multimodal collaborative perception framework (MCPF). Transceiver side images are compressed into embeddings by convolutional neural networks (CNNs), while CIR acquires embedded representations through 1 dimensional convolutional layers combined with LSTM modules. Adaptive weight allocation dynamically fuses these cross-modal features, enabling a lightweight fully-connected classifier to distinguish NLOS conditions. To optimize cross-modal interactions, this paper further designs a tailored composite loss function specifically for the multimodal architecture. Experimental validation on a field-collected dataset spanning eight LOS/NLOS scenarios demonstrates that MCRF reaches 94.31%, and significantly outperforms the classical single modal methods (e.g., kurtosis, LSTM, CNN-LSTM) by 10.59-26.53 percentages.

This paper proposes a novel equalizer for a hybrid-MIMO system combining both linear and nonlinear magnitude-only radio-frequency chains. A dropout-and-constellation-deletion parametric bilinear generalized approximate message passing (DCDP-BIG-AMP) algorithm is developed, and then applied for joint channel and data estimation (JCDE) in the hybrid-MIMO. Compared to parametric bilinear generalized approximate message passing (P-Bi-GAMP), DCDP-BIG-AMP randomly selects only a portion of variables to be updated and deletes constellation points which are changed slightly during the DCDP-BIG-AMP iterations. Consequently, the computational complexity is decreased significantly with the cost of minor performance degradation based on our experimental results. Its core operation is cyclic convolution, which is accelerated by fast Fourier transform (FFT). Subsequently, it is suitable for parallel hardware implementation. Moreover, the expectation maximization (EM) framework is combined to learn the unknown prior distribution parameters of channel. Simulation results show that DCDP-BIG-AMP successfully exploits the nonlinear magnitude measurements, and enables hybrid-MIMO to outperform the traditional MIMO on communication performance and energy-efficiency. DCDP-BIG-AMP based JCDE alternatively enhances the channel estimation (CE) and multiuser detection (MUD) performances, and converges quickly after only about 5 iterations. The dropout and constellation deletion mechanisms well balance between computational complexity and performance degradation.

This correspondence discusses recognition of traffic modes and behavior patterns based on Global Navigation Satellite System (GNSS) data. The traffic modes (e.g., walk, car, train, etc.) are firstly inferred, and then their behavior patterns (e.g., left-turn, right-turn, turn-around, etc.) are further identified. Because both traffic modes and behavior patterns are strongly influenced by geographical circumstances, their recognitions are enhanced by geographical layer information (e.g., building, road, water, etc.). At one specific GNSS point, its surrounding area is uniformly sliced as grids, and the probabilities for grid centers belonging to six different geographical layers are calculated based on whether these centers lie inside or outside of the minimum rectangles containing polygons indicating different geographical objects. Finally, the six-dimensional probability matrix is processed and compressed as a geographical information vector by the convolutional neural network (CNN). The latter is then combined with kinematic metrics such as velocity, acceleration, and moving direction from GNSS data, and serially input into a long short-term memory (LSTM) network to predict traffic modes and behavior patterns. Experimental results validate that the geographical information does enhance the performances of two recognition tasks. The CNN+LSTM framework retains the powers of CNN and LSTM, and outperforms classical machine learning algorithms.

This paper proposes a nonlinear MIMO scheme named as halved-phase only MIMO (HPO-MIMO), whose base station (BS) is equipped with multiple HPO radio-frequency (RF) chains extracting$\pi $-periodic phase measurements from RF signals directly by phase detectors, and single classical RF chain sampling the combination of all the RF signals from HPO-RF chains. Compared to MIMO, HPO-RF receivers are simplified significantly, and have much lower power consumption and fabrication cost. Two types of channel estimators and multiuser detectors are developed for HPO-MIMO from the perspectives of numerical optimization and Bayesian inference. Firstly, because$\pi $-periodic phases provide tan-relationships between the Inphase (I) and Quadrature (Q) components, the channel estimation (CE) and multiuser detection (MUD) problems are resolved by minimizing$l_{2}$-norm cost functions with unit-norm constraint on the recovered vector. Their solutions are obtained by deriving the eigenvector corresponding to the minimum eigenvalue through shifted power method (SPW). Secondly, the CE and MUD problems are categorized as generalized linear mixing problems under (un-)quantized$\pi $-periodic phase measurements, and then handled by generalized approximate message passing (GAMP), where closed-form solutions for mean-and-variance messages involving$\pi $-periodic phases are derived. Finally, magnitude and$\pi $-phase ambiguities persisting in signal recovery are removed based on the complex-valued full measurements from the single classical RF chain. Simulation results show that GAMP-type algorithms outperform SPW-type ones, and handle nonlinear phase quantization losses. 16-sector quantized HPO-MIMO could work as well as its un-quantized correspondence. HPO-MIMO reserves the MIMO advantages, but is more energy-efficient than the latter.

This paper researches the cooperative localization in wireless sensor networks (WSNs) with$2\pi /\pi $-periodic angle-of-arrival (AOA) measurements. Two types of localizers are developed from the perspectives of Bayesian inference and convex optimization. When the orientation angles are known, the positioning problem is resolved by a phase-only generalized approximate message passing (POG-AMP) algorithm with importance sampling mechanism. From the perspective of convex optimization, the positioning problem under$2\pi /\pi $-periodic AOAs is converted as a least square (LS) problem and then resolved by the gradient-descent/projected gradient-descent method named as Type-I LS localizer. When the orientations are unknown, expectation-maximization (EM) mechanism is introduced into the POG-AMP localizer, where node positions and orientations are alternatively updated through exchanging their statistical confidences. Type-II LS localizer is constructed by alternatively executing Type-I LS and a maximum-likelihood (ML) estimator of orientation. Cramér-Rao lower bounds (CRLBs) are derived for the proposed localizers. Simulation results validate that the proposed AMP-type and LS-type localizers outperform existing localizers, AMP-type localizers successfully handle nonlinear quantization losses, and EM-framework and ML estimator handle unknown orientation problem. AMP-type localizers outperform LS-type ones, and can approach to the CRLBs even under high noise contaminations.

In this paper, positioning performances of vehicular ad-hoc networks (VANETs) are significantly enhanced by cooperative localization (CL) based on the three dimensional (3D) angle-of-arrival (AOA) measurements. Localizers are proposed based on generalized approximate message passing (GAMP), which works under (un-)known Euler rotation angles, and π/2π-periodic azimuth AOAs. Firstly, when the Euler angles are known, GAMP localizer is obtained by categorizing the AOACL problem as a generalized linear mixing one, and then resolved by GAMP, whose mean-and-variance messages involving 3D AOAs are evaluated by importance sampling technique. Secondly, when the Euler angles are unknown, the expectation maximization (EM) framework is combined with GAMP, where vehicle positions and Euler angles are alternatively updated through one-step GAMP iteration and maximizing conditional probability distribution functions (pdfs) expected over hybrid variables, which are obtained by subtracting the xyz position of the vehicle receiving measurements with the position of its pairing neighbor, respectively. Simulation results validate that the proposed localizers outperform existing localizers. Positioning accuracy can be significantly enhanced compared with global navigation satellite system (GNSS).

In this paper, a three-dimensional universal cooperative localizer (3D UCL) is proposed for vehicular ad-hoc networks (VANETs) in 3D space under varied types of ranging measurements including time-of-arrival (TOA), received signal strength (RSS), angle-of-arrival (AOA), and Doppler frequency. Its core idea is to exploit generalized approximate message passing (GAMP) to resolve the 3D cooperative positioning problem after converting it as a generalized linear mixing problem. Unfortunately, the positioning performance of 3D UCL is severely degraded by the inaccurate ranging measurements from the non-line-of-sight (NLOS) links. Therefore, a 3D geographical information enhanced UCL (3D GIE-UCL) is developed by combining 3D UCL with a NLOS identification mechanism assisted by geographical information. Finally, 3D UCL is accelerated by graphics processing unit (GPU) parallelization, particle reduction and message censoring. 3D GIE-UCL is accelerated by particle reduction and anchor upgrading. Simulation results validate state-of-the-art positioning performances and cooperative gains of both 3D UCL and 3D GIE-UCL after comparing them with existing cooperative localizers. 3D GIE-UCL approaches to its performance upper bound provided by its correspondence with oracle link-type information. 3D UCL and GIE-UCL show 241× and 3.3× speedup after adopting the acceleration techniques, respectively.

The spatio-temporal sparsity (STS) is exploited to improve the channel estimation (CE) for nonlinear multiple input multiple output (NL-MIMO), whose base station (BS) acquires only phases-or-magnitudes of the received complex signals through low-power and low-cost phase/envelope detectors. The sparse CE problem is formulated as the generalized linear mixing one and resolved by a modified generalized vector approximate message passing (GVAMP) algorithm with expectation maximization (EM) mechanism. The prior distribution of sparse channel is modeled as Bernoulli Gaussian-mixture (BGM). Consequently, NL-MIMO channel responses and unknown parameters including BGM parameters and noise variance are updated by the GVAMP and EM procedure alternatingly. A gradient descent (GD) method is proposed for EM update of noise variance in NL-MIMO, where closed-form solution is missed due to the complex formulas of likelihood under phase/magnitude observations. Monte Carlo integration technique is exploited to numerically derive the gra-dient of the noise variance under phase/magnitude observations. Simulation results show that the transmission power and pilot length are significantly saved after exploiting the STS, and both the EM-GVAMP and GD estimators are validated.

This paper researches the cooperative localization in wireless sensor networks (WSNs) with $2\pi/\pi$-periodic angle-of-arrival (AOA) ranging measurements. When the orientation angles of the antenna arrays at WSN nodes are known, a ranging link loss is defined based on the tan-relationships between AOA observations and xy-minus coordinates of two neighboring nodes. Subsequently, the positioning problem under $2\pi$-periodic AOAs is converted as a convex optimization problem about minimizing total ranging-link loss through optimizing agent positions, which is resolved by the gradient-descent (GD) method. Under $\pi$-periodic AOAs, additional 0/1 integers are introduced to indicate the front-or-back impinging directions. By relaxing 0/1 integers as continuous variables within $[0,1]$, the positioning problem is relaxed as a nonconvex optimization one about minimizing total link loss over the agent positions and indicating variables, which is solved by the projected GD (PGD) method. Finally, Type-I least-square (LS) localizer is developed for WSNs with both $2\pi$ and $\pi$-periodic AOAs. When the orientation angles are unknown, Type-II LS localizer is developed by combining Type-I LS localizer with a maximum-likelihood (ML) orientation estimator, which alternatively updates agent positions and orientation angles. Simulation results validate that the proposed LS-type localizers outperform existing localizers.

Cooperative positioning is promising to provide stable and precise location information for vehicular ad-hoc networks (VANETs). However, its performance is severely degraded by the non-line-of-sight (NLOS) ranging measurements due to blockages from buildings and vehicles. In this paper, a three-dimensional (3D) geographical information enhanced cooperative localizer (3D GIE-CL) is proposed for VANETs with time-of-arrival (TOA) ranging measurements from mixed lineof- sight (LOS) and NLOS links. At every iteration, it firstly judges the NLOS due to buildings (NLOSb) and NLOS due to vehicles (NLOSv) based on geographical information and current estimations of vehicles positions. Secondly, after removing the judged NLOS measurements, it updates the positions estimations by a 3D generalized approximate message passing (GAMP) CL, which handles the cooperative positioning problem by treating it as a generalized linear mixing one, and then resolving the latter by GAMP with importance sampling mechanism to evaluate the mean-and-variance messages involving TOA measurements. Finally, 3D GIE-CL will iterate between the above NLOS identification and 3D cooperative localization until convergence is reached. Simulation results show that 3D GIE- CL handles the NLOS problem successfully. It provides the state-of-theart positioning performance and approaches to its performance upper bound under oracle link-type information.

This paper proposes a halved-phase only multipleinput- multiple-output (HPO-MIMO) whose base station (BS) acquires pi-periodic phases of complex envelope signals by phase detectors. Because phases are directly extracted from the received radio frequency (RF) signals, the RF receivers of HPO-MIMO are simplified significantly, and enjoy much lower power consumption and fabrication cost in comparison to their correspondences in MIMO. However, existing MIMO baseband algorithms will not work in HPO-MIMO because the latter losses magnitude information completely and its phase observations suffer from pi-ambiguities. Practical baseband algorithms are developed for HPO-MIMO by firstly categorizing the multiuser detection (MUD) and channel estimation (CE) problems as generalized linear mixing ones under pi-periodic phase measurements, and then resolving the latter by a modified generalized approximate message passing (GAMP). In GAMP, the mean-variance messages involving phase observations are numerically updated by the importance sampling technique. Scale-and-polarity ambiguities among the CE and MUD results are removed based on the measurements from a conventional RF chain. Simulation results validate the effectiveness of the proposed algorithms, and compare HPO-MIMO with existing MIMO schemes thoroughly.

Cooperative localization (CL) based on the angle-of-arrival (AOA) measurements is a promising positioning technique for the wireless sensor network (WSN). This is because CL reaches high localization precision and robustness by exploiting the relative ranging measurements among the agents. In addition, the AOA interpretation involves no propagation parameters, and its acquisition does not require strict time-synchronization among the WSN nodes. In this paper, we firstly categorize the AOA-CL problem as a generalized linear-mixing problem under the phase-only measurements, and then resolve it by our developed phase-only generalized approximate message passing (POG-AMP) algorithm. The POG-AMP localizer is warm-started by developing an incremental localizer assisting by anchor-connectivity and region-boundary constraints. It calls for sparse matrix-vector multiplications (MVMs) as its most complex operations, and can be implemented in a distributed manner. Therefore, it has low computational complexity, and is suitable for WSN hardware implementation. Simulation results validate the state-of-the-art performance and cooperation gains of the POG-AMP localizer.

Cooperative localizer is a potential positioning technique for vehicular ad-hoc networks (VANETs). However, it would suffer from the non-line-of-sight (NLOS) problems widely existing in VANETs. This letter proposes a geographical information enhanced cooperative localizer (GIE-CL) for VANET with time-of-arrival (TOA) measurements. It iterates between NLOS identification and extended generalized approximate message passing (EGAMP) motivated cooperative positioning. A region sampling method is developed to identify NLOS measurements based on geographical information and current vehicle position estimations. Subsequently, the detected NLOS measurements are removed and the EGAMP localizer is activated to re-estimate the vehicle positions. The above-mentioned iteration will be terminated until convergence is reached. Initial positions are provided by Global Navigation Satellite System (GNSS). Simulation results show that GIE-CL can handle the NLOS problem, and approach to its performance upper bound provided by the case with known NLOS/LOS link-type information. Compared to EGAMP localizer, the positioning accuracy of GIE-CL is improved by eight times when the allowed localization error is less than 5 m.

This letter proposes a universal cooperative localizer (UCL) for wireless sensor network (WSN) with varied types of ranging measurements including time-of-arrival, radio signal strength, angle-of-arrival, and Doppler frequency. By representing the WSN node positions as complex numbers, the localization problem is converted as a generalized linear mixing problem, and then resolved by generalized approximate message passing with the importance sampling mechanism. The computational complexity of UCL is two orders lower than that of the belief propagation localizer. Simulation results validate that UCL outperforms most mainstream localizers, and cooperation significantly enhances the positioning accuracy.

This paper proposes a multiuser magnitude-only (MO-)MIMO, whose base station acquires quantized magnitudes of the complex baseband signals through envelop detectors and low-resolution ADCs. Consequently, MO-MIMO enjoys much lower circuit power and cost in comparison with the conventional MIMO. Because the phase information is unavailable, all the existing MIMO baseband algorithms cannot be applied into MO-MIMO. Therefore, two types of channel estimators and multiuser detectors are constructed by first categorizing the channel estimation and multiuser detection problems as a quantized phase retrieval (PR) problem, and then solving the latter by developing two methods under the framework of generalized approximate message passing (GAMP). The first method directly applies GAMP to solve the quantized PR problem by exploiting the probability relationships between the quantized magnitude measurements and unknown complex signals. The second method iterates between the missing phase estimation and signal recovery, where the latter calls for GAMP to handle a linear mixing problem with quantized observations. The developed estimators and detectors call for matrix-vector multiplications and nonlinear function calculations as the most complex operations, handle the nonlinear quantization loss specially, and exploit the signal prior probability distributions. Finally, their effectiveness is validated experimentally.

This paper proposes a MIMO system where the base station (BS) acquires quantized phase-only (PO) measurements of the complex baseband signal by our introduced stage-wised phase quantizer. PO-MIMO requires only one-bit ADCs for data sampling, so it successfully overcomes the ADC bottleneck that appears when the signal bandwidth is extremely wide. We construct a PO generalized approximate message passing (POG-AMP) algorithm for solving the linear mixing problem with quantized or unquantized phase measurements. POG-AMP has low computational complexity, exploits the signal prior statistical distribution, and handles the nonlinear distortions exerted on the measurements (e.g., losing magnitude and quantization). Then, POG-AMP is successfully applied to construct practical channel estimator and multiuser detector for PO-MIMO. Numerical results show that the POG-AMP estimator (POG-AMPE) and POG-AMP detector (POG-AMPD) are robust to the phase-quantization loss. POG-AMPE acquires high-quality channel side information at the receiver (CSIR), and POG-AMPD is robust to the CSIR errors when the BS antennas are massive enough. By introducing moderately more BS antennas, PO-MIMO with phase measurements even performs similarly to MIMO with full measurements containing both magnitude and phase. In order to maximize the transmit energy-efficiency, the lengths of the channel training sequences should be gradually increased with the increase of the channel coherence time. Antenna correlations at the BS degrade the convergence and bit-error rate performances of POG-AMPD, but can be handled by the analog spatial filtering technique.

This paper proposes a massive multiple-input multiple-output (MIMO) with low-precision in-phase quadrature-phase (IQ) (de-)modulators and low-resolution analog-to-digital convertors, and develops uplink baseband algorithms for the proposed MIMO system. The IQ imbalance parameters are recovered by a two-stage method. At the first stage, the base station (BS) antennas acquire phase-shifted versions of their imbalance parameters after receiving some predefined training sequence emitted by a specific transceiver that is near to the BS and shares flat-fading channels between itself and the BS. At the second stage, the phase shifts from the first stage are estimated by the specific transceiver through receiving a common sequence transmitted by the BS antennas alternatively. In order to jointly handle the IQ imbalance and nonlinear quantization loss, the IQ-imbalanced multiuser (MU)-MIMO with single-antenna users is first converted into an IQ-balanced MU-MIMO with dual-antenna users. Then, two kinds of channel estimator and multiuser detectors are constructed for the equivalent IQ-balanced system based on the spectral projected gradient method and vectorized message passing de-quantization algorithm. Our numerical results validate the effectiveness of the above baseband algorithms, and show that the proposed massive MIMO is more energy-efficient than the conventional one when the channel coherence time is long enough.

This paper proposes a multiuser MIMO system with both full measurements and one-bit magnitude observations, which are acquired by several linear inphase-and-quadrature (IQ) structured radio frequency (RF) chains and massive one-bit envelope chains, respectively. The total circuit power and cost are not increased significantly, since the added one-bit envelope chains have low power and low cost. Channel side information on the one-bit envelope chains is acquired by sharing the IQ-structured RF chains and executing a channel calibration operation. Two multiuser detectors are constructed based on the semidefinite relaxation (SDR) and approximate message passing (AMP). The one-bit magnitudes are interpreted as inequality constraints in the SDR detector, and exploited in a Bayesian manner by the AMP detector. Simulation results show that the one-bit magnitude measurements bring about high MIMO multiplexing and diversity gains, and decrease the transmission power. With the increase of the channel coherence time, more one-bit envelope chains are prone to be equipped, and one-bit magnitude-aided MIMO becomes more and more spectral-and-energy-efficient than the conventional MIMO.

This correspondence researches the multiuser detection (MUD) in a new massive MIMO system, where the base station (BS) acquires the quantized amplitude-only (AO) measurements of the complex baseband signals through the envelop detection and low-resolution analog-to-digital convertor (ADC) sampling. Cumbersome frequency mixers are removed and low-resolution ADCs can easily reach the Giga-sample-per-second (GSPS) rate, so massive AO-MIMO has low power and low cost, and is especially suitable for millimeter-wave (mmWave) communication. A practical multiuser detector is constructed based on the Wirtinger-flow (WF) gradient method for phase retrieval (PR), and the message passing de-quantization (MPDQ) algorithm for the quantized compressed sensing (CS). It iterates between the missing phase estimation and signal reconstruction, and involves only matrix-vector multiplication as its complex operation. Simulation results validate the effectiveness of the proposed detector, and show that massive AO-MIMO can even perform similarly as the classical massive MIMO by adopting relatively more BS antennas.

This paper proposes a new large-scale antenna system (LSAS) where the base station (BS) is equipped with several classical radio-frequency (RF) chains but massive one-bit integrated information and energy receivers (Rxs). Multiple users with multiple antennas are served simultaneously. This new LSAS is much more energy efficient than massive MIMO because it requires much less classical power-hungry RF chains, and its massive integrated Rxs can harvest new energy. For the uplink multiuser detection (MUD), we develop a practical generalized approximate message passing detector (GAMPD), which exploits the signal prior probability distribution, and the hybrid BS measurements (full ones with both amplitudes and phases from the classical Rxs, and the one-bit amplitude measurements from the integrated Rxs) in a Bayesian manner. GAMPD involves matrix-vector multiplications, small-scale matrix inversions, and parallelized nonlinear operations, so it is suitable for hardware implementation. Our numerical results indicate that the proposed LSAS retains the advantages of massive MIMO, and can even perform similarly as the latter after introducing sufficient number of integrated Rxs.

In this paper, we research the multiuser detection (MUD) in a new massive MIMO with quantized phase-only (PO) measurements. Because the phase quantization can be implemented by one-bit ADCs, PO-MIMO overcomes the ADC bottleneck that appears when the signal bandwidth reaches multi-giga Hertz. However, the new problem is how to complete the uplink multiuser detection (MUD) based on the quantized PO measurements. In this paper, it is firstly solved by the sum-product-algorithm (SPA). Based on central limit theorem (CLT) and Taylor expansions, the SPA detector (SPAD) is simplified as a new PO generalized approximate message passing detector (POG-AMPD). POG-AMPD can exploit the quantized phase measurements and signal prior probability distribution in a Bayesian manner. It involves matrix-vector multiplications, and some parallelized non-linear operations, so it is suitable for hardware implementation. Numerical results indicate that POG-AMPD approaches to the maximum likelihood (ML) detection performance. By adopting moderately more BS antennas, massive PO-MIMO can even performs similarly as the conventional massive MIMO with full measurements.

In this paper, we research the multiuser detection (MUD) in a new Spatial Modulation Multiple-Input-Multiple-Output (SM-MIMO) system, whose Base Station (BS) is equipped with massive Radio-Frequency (RF) chains with low-resolution Analog-to-Digital Convertors (ADCs), and the User Equipments (UEs) have multiple Transmit Antennas (TAs) but single RF chain. In the uplink, UEs transmit their data by the Cyclic-Prefix Single-Carrier (CP-SC) SM technique. The key is how to design practical MU detectors to handle the severely quantized measurements and antenna correlations. Coherent detection is focused, so a Least-Square (LS) channel estimator is designed to acquire the the channel side information (CSI) at the Receiver (CSIR). Then, we firstly solve the MUD problem by the Sum-Product-Algorithm (SPA) on a clustered factor graph (FG) whose variable nodes correspond to the transmitted vectors from the UEs. Next, based on the Central Limit Theorem (CLT) and Taylor expansions, the SPA detector (SPAD) is simplified as a new low-complexity Message Passing De-Quantization Detector (MPDQD), which exploits both the structured sparsity and prior probability distribution of the transmitted signal. By utilizing the clustering technique, damping mechanism, and Analog Spatial Filtering (ASF), the robustness of MPDQD is improved significantly. Simulation results show that MPDQD outperforms the linear detectors, works steadily under strong channel correlations, and even performs similarly as its counterpart in the un-quantized SM-MIMO.

In this work, we research the multiuser detection (MUD) in a new Single Carrier (SC) massive Spatial-Modulation-Multiple-Input-Multiple-Output (SM-MIMO) system, where the Base Station (BS) is equipped with massive antennas connecting to independent Radio-Frequency (RF) chains designed by low-resolution Analog-to-Digital Converters (ADCs). Different User Equipments (UEs) possess multiple Transmit Antennas (TAs) but only one RF chain, and transmit data-bits to the BS by Cyclic-Prefix Single-Carrier (CP-SC) SM. We construct a new Message Passing De-Quantization Detector (MPDQD) based on the MPDQ algorithm in Compressed Sensing (CS). MPDQD exploits both the special structures (e.g., sparsity) and prior probability distribution of the transmitted signal, so it reaches superior detection performances. It involves the parallelized matrix-vector multiplication as the most complex operation, so it has low complexity and is suitable for hardware implementation. Simulation results show that MPDQD outperforms the linear detectors, and can approach the performance of the un-quantized system even when ADC resolution is low (e.g., 4bits).

In large-scale Multiple Input Multiple Output (MIMO), the number of Base Station (BS) antennas reaches tens even hundreds. If so many radio frequency front-ends adopt the classical receivers, their cost and power consumption during the receiving mode would increase quickly. Consequently, we exploit one-bit Analog Digital Convertor (ADC) to design low-cost and low-power software defined radio receivers for the BS. However, the new problem is how the BS fulfills Multiuser Detection (MUD) based on the one-bit baseband data. Based on the Message Passing De-Quantization (MPDQ) algorithm, an iterative multiuser detector is constructed. Its convergence properties is analyzed by the State Evolution (SE), and then its complexity is proved to be one order of magnitude smaller than that of the Minimum Mean Square Error (MMSE) detector. Simulation results show that the proposed detector outperforms MMSE. Finally, the proposed detector with matrix-vector multiplications is suitable for hardware implementation.

In large-scale Multiple-Input-Multiple-Output, the Base Station (BS) is equipped with a large-size antenna array containing tens even hundreds Radio Frequency channels. If so many RF front-ends adopt the classical receivers, their cost and power consumption during the receiving mode would increase quickly. Therefore, we exploit low-resolution Analog-to-Digital-Convertor to design low-power and low-cost software defined radio receivers for the BS. However, the new problem is how the BS fulfills channel estimation and Multiuser Detection (MUD) under low-resolution quantization. In this paper, first, least square method is used for channel estimation, and a robust Maximum Likelihood (ML) MUD problem is constructed to take into account the channel estimation errors. Second, an iterative multiuser detector is constructed by relaxing the ML MUD problem as a convex optimization problem and then solving the convex problem through the nonmonotone spectral projected gradient method. Compared with the Minimum Mean Square Error (MMSE) detector, the proposed detector has lower computational complexity, and is more suitable for hardware implementation. Simulation results show that it also outperforms MMSE.

In large-scale Multiple-Input-Multiple-Output (MIMO) downlink, one state-of-the-art precoder significantly reduces the Peak-to-Average Power Ratio (PAPR) of the emitted signals from the Base Station (BS) antennas through minimizing the ℓ∞-norm of the precoded signals (generated by the precoder and transmitted by the BS RF frontends). In this paper, its computational complexity is analyzed based on the random matrix theory, and its two variants are constructed as follows. First, the ℓ∞-norm is approximated by a high order ℓp-norm (e.g. ℓ16), and then the downlink precoding problem is solved by classical Gradient Descent (GD) method. Second, the ℓ∞-norm is removed directly after additional box constraints are exerted on the elements of the precoded signals. Then the precoding problem is solved by spectral GD algorithm. Simulation results show that the above three convex-optimization based precoders can reduces the PAPR by more than 11.5dB compared to conventional precoding schemes.

In this paper, an iterative multiuser detector is proposed for uplink large-scale Multiple Input Multiple Output (MIMO). It is developed based on the Generalized Approximate Message Passing (GAMP) algorithm, and its convergence properties are analyzed by a one-dimensional iteration termed as State Evolution (SE). The SE analysis proves that the complexity of the GAMP detector (GAMPd) is one order of magnitude smaller than that of the Minimum Mean Square Error (MMSE) detector. Simulation results show the GAMPd performs similarly to MMSE at least, and even outperforms the latter when the number of BS antennas is not much larger than the number of users. In addition, the GAMPd with matrix-vector multiplications is suitable for hardware implementation.

Approximate Message Passing (AMP) simplified from Loopy Belief Propagation (LBP), is an important algorithm for sparse signal reconstruction in Compressed Sensing (CS). To improve the performance of current AMP algorithms, a weighted-damped AMP algorithm (WDAMP) is derived from a weighted version of BP that adopt probability damping technique. Simulation results show that WDAMP outperforms normal AMP for both 1-D and 2-D signal reconstruction. For 1-D signal reconstruction, probability damping brings most of the improvement. For 2-D signal reconstruction, weighting technique makes the major contribution. In summary, WDAMP outperforms conventional AMP.