This paper proposes a low-complexity, machine learning (ML)–aided channel denoising framework that applies element-wise filtering in the beam–delay domain to enhance multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) channel estimation. In new radio (NR), demodulation reference signals (DMRSs) enable direct estimation of a subset of the channel state information (CSI) and the remaining CSI is reconstructed by interpolation. However, the noise from reference-based direct estimation can degrade overall accuracy. To address this, we introduce two compact neural network architectures: a shallow single-stream model and a dual-stream factorized model, both built from linear layers, complex domain rectified linear unit (cReLU) activations, and a custom normalization-threshold function. Under the 3GPP UMi channel model, extensive simulations demonstrate that our denoisers outperform classical beam–delay thresholding and a conventional convolutional neural network (CNN)-based method in normalized mean square error (NMSE) and computational cost. Specifically, the proposed designs reduce floating-point operations (FLOPs) by over 95% compared to the CNN benchmark while achieving more than 10% relative NMSE improvements.

Channel denoising plays a critical role in enabling accurate channel estimation for modern multiple-input multiple-output (MIMO)–orthogonal frequency division multiplexing (OFDM) systems. As antenna counts and frequency bands proliferate, channel impulse responses become increasingly complex, challenging conventional denoising methods. Motivated by the success of data-driven techniques, we introduce a novel, fully complex-valued transformer architecture tailored for beam–delay domain channel denoising. Key innovations include an inverse exponential positional embedding that avoids corrupting dominant delay taps and an encoder-only design that streamlines one-to-one mapping from noisy to clean channel matrices. The network is trained in a supervised fashion to minimize mean square error (MSE) loss function. Simulation results using 3GPP Urban Micro channel model at 3.5 GHz carrier frequency demonstrate that the proposed framework reduces mean estimation error compared to the legacy threshold-filter method and recent state-of-the-art machine learning (ML)-based denoisers, by a significant margin at 44% and 33%, respectively.

In this paper, we propose a reduced-complexity detector for non-coherent communications with the Grassmann constellation, which enables joint channel and data estimation. Here, we employ deep learning techniques with the aim of reducing computational complexity while maintaining nearly the same channel estimation accuracy as the conventional detector. The conventional maximum likelihood detection is a discrete optimization problem, with complexity increasing exponentially with respect to the transmission rate. In our approach, the Grassmann constellation is detected using a neural network model trained with random signal-to-noise ratios and the corresponding received signals as inputs, and the detected codeword is used for channel estimation. Our simulations demonstrate that the channel estimation accuracy can be maintained with lower complexity compared to the conventional detector, at the cost of a slightly degraded symbol error rate performance. It was also found that the detection complexity can be reduced when the number of receive antennas exceeds four, which is practically relevant.

In this paper, we analyze the performance of the data-carrying reference signal (DC-RS) in time-varying channels. In such scenarios, the spectral efficiency may be reduced because more reference signals need to be transmitted frequently to maintain high channel estimation accuracy. DC-RS has the potential to boost spectral efficiency, which conveys additional data with noncoherent detection, but it has not been analyzed in realistic time-varying channels. By regarding a channel coefficient varying with first-order autoregressive model as an additive independent Gaussian noise for each time slot, we derive the average mutual information of DC-RS. Although the time-varying nature induces performance penalty in general, our numerical simulations demonstrate that the spectral efficiency improves even in rapidly varying channels compared to the case with conventional reference signals, and this trend remains valid upon increasing the mobile speed from 0 to 300 km/h.

In wireless networks, frequent reference signal transmission for accurate channel reconstruction may reduce spectral efficiency. To address this issue, we consider to use a data-carrying reference signal (DC-RS) that can simultaneously estimate channel coefficients and transmit data symbols. Here, symbols on the Grassmann manifold are exploited to carry additional data and to assist in channel estimation. Unlike conventional studies, we analyze the channel estimation errors induced by DC-RS and propose an optimization method that improves the channel estimation accuracy without performance penalty. Then, we derive the achievable rate of noncoherent Grassmann constellation assuming discrete inputs in multi-antenna scenarios, as well as that of coherent signaling assuming channel estimation errors modeled by the Gauss-Markov uncertainty. These derivations enable performance evaluation when introducing DC-RS, and suggest excellent potential for boosting spectral efficiency, where interesting crossings with the non-data carrying RS occurred at intermediate signal-to-noise ratios.

In this paper, we propose an accurate channel estimation scheme for orthogonal time frequency space modulation, in which we embed multiple superimposed pilots (SPs) instead of a single SP used in the conventional scheme. Multiple SPs are used to mitigate data-pilot interference efficiently and improve the accuracy of channel estimation. Our numerical simulations demonstrate that the proposed scheme outperforms the conventional SP-based scheme in terms of bit error rate and normalized mean squared error of channel estimates, indicating the potential for further improvement in spectral efficiency.

In this paper, a novel training beam sequence design for multiuser millimeter wave tracking systems is proposed. For each receiver, a single-path channel model is firstly investigated, where we introduce a maximum a posteriori (MAP) criterion to estimate the time-varying angle of departure (AoD), followed by an extended Kalman filter to update the stale complex path gain. We then employ training beam sequence design to minimize the estimated AoD’s average mean squared error (AMSE), which however has no explicit expression. We firstly derive a closed-form upper bound for the AMSE and then simplify this upper bound into a tractable form, based on which a nonlinear optimization problem (NLP) is formulated. By solving this NLP optimally using its corresponding Karush-Kuhn-Tucker conditions, we obtain an efficient training beam sequence. The proposed MAP criterion and its associated training beam sequence design are further extended to multi-path scenarios, where a joint estimation of the multiple paths is firstly discussed, followed by a sequential estimation as a low-complexity alternative. Numerical results demonstrate the superiority of our proposed scheme over the existing benchmark methods, especially in the case when the receivers’ channels change rapidly.

We consider computation-offloading for Internet-of-things (IoT) applications in multiple-input-multiple-output (MIMO) cloud-radio-access-network (C-RAN). Specifically, the computational tasks of the IoT devices (IoTDs) are offloaded to a MIMO C-RAN, where a MIMO radio resource head (RRH) is connected to a baseband unit (BBU) through a capacity-limited fronthaul link, facilitated by the spatial filtering and uniform scalar quantization. We formulate a computation-offloading optimization problem to minimize the total transmit power of the IoTDs while satisfying the latency requirement of the computational tasks. To obtain a feasible solution for the non-convex problem, firstly the spatial filtering matrix is locally optimized at the MIMO RRH. Subsequently, leveraging the alternating optimization framework for joint optimization on the residual variables at the BBU, the baseband combiner, the optimal resource allocation and the number of quantization bits are obtained through the minimum-mean-squared-error (MMSE) metric, the successive inner convexification method and the line-search method, respectively. As a low-complexity approach, we apply a supervised deep learning (DL) method, which learns from the solutions obtained with our proposed algorithm. In addition, the deep transfer learning is adopted to adjust the neural network in dynamic IoT systems. Numerical results validate the effectiveness of the proposed optimization algorithm and the learning based methods.

In this paper, we propose the hybrid-precoding design that alleviates the performance loss caused by beam-misalignment in the mmWave multi-user communication systems. To this end, we firstly design the beam-misalignment aware fully-digital precoders for two distinct scenarios. First, for a base-station (BS) with full estimated channel-state-information (CSI), the minimum-mean-squared-error metric incorporating the `error-statistics' of the beam-misalignment error is used to analytically derive a closed-form expression for the fully-digital precoder, which maximizes the array gain while suppressing the inter-user interference for each user-equipment (UE). Second, for a BS which can only acquire partial estimated CSI, a min-max non-convex optimization is considered to obtain the fully-digital precoder, which minimizes the maximum loss in the array gains of the expected beam-misalignment `error-range' over the UEs while cancelling the inter-user interference. Subsequently, we propose the hybrid-precoding design that approximates the fully-digital designs based on the gradient-projection method, which is mathematically proven to converge to an approximate local solution with further reduced complexity compared to the state-of-the-art algorithms. Finally, the proposed hybrid-precoding design is further extended to the wideband mmWave communication systems. Numerical results show that the proposed hybrid-precoding design can effectively alleviate the performance degradation incurred by the beam-misalignment.

River interventions disturb the natural flow regime of channel. The impacts generated from river interventions is categorized based on the alteration of fluvial system including channel planform and ecology. The Brahmani River, one of the major peninsular river holds a significant importance for three major states of eastern India. This river has been subjected to periodic flooding events damaging high value property. Rengali multi-purpose river valley project (since 1985), located along middle reach of Brahmani, influence the downstream stretch behavior. A planform variation of increased braiding pattern is observed for downstream river. Interaction of reservoir-catchment along with planform variation assessment embraces a substantial importance for geologists and engineers. This research article aims to quantify variation in channel planform for pre-Rengali and post-Rengali time period along a selected reach integrating Sinuosity Index (SI) and Braid-Channel Ratio (BR) parameters using remote sensing techniques.

This paper proposes the revamp of architecture of an eNodeB (eNB) for 5G cellular communication networks capable of opening up the crunched spectrum resources. We have developed a novel priority driven Resource Block (RB) handoff algorithm for Full-Duplex Multi-User MIMO communication, where the eNB creates beams for establishing communication links. The use of Full-Duplex operation enables simultaneous in-band uplink and downlink operation without sacrificing temporal resource in each link. Space-division multiple access (SDMA) technique has been used to dynamically create a beam towards the intended mobile users as well as track them through beam steering. The possibility of co-channel interference (CCI) due to mobility of users is eliminated by the introduction of RB handoff mechanism. This allows sharing of a RB or RBs by multiple users. In the proposed architecture, along with coexistence of multiple mobile UEs in same RBs diversity gain is introduced both in uplink and downlink. The work includes related simulations showing the conceptual validity of the proposed architecture to revamp. The proposed eNB model leads to improved QoS and a significant decrease in cellular spectrum requirement worth millions of dollars.