In-X subnetworks (SNs) encounter significant challenges in achieving hyper-reliable low-latency communication (HRLLC), particularly in hyper-dense deployment scenarios. These challenges stem from rapid and dynamic variations in interference caused by mobility, dynamic channel statistics, and varying traffic patterns. Predictive interference management plays a crucial role in meeting these extreme requirements, by enabling proactive and reliable resource allocation to prevent performance degradation. To address these challenges, we propose a probabilistic interference prediction technique using an inverted quantile transformer (iQTransformer) to learn interference dynamics effectively and capture tail statistics. It predicts interference for multiple sensor-actuator (SA) pairs in the SN with minimal disparity by leveraging their own interference dynamics alongside those of other active SA pairs. We integrate extreme value theory (EVT) with the iQTransformer to handle rare and extreme interference events. Results demonstrate that our proposed method outperforms baseline approaches in terms of prediction accuracy, achieving near-optimal target reliability.

In this paper, we explore a multi-task semantic communication (SemCom) system for distributed sources, extending the existing focus on collaborative single-task execution. We build on the cooperative multi-task processing introduced in [1], which divides the encoder into a common unit (CU) and multiple specific units (SUs). While earlier studies in multi-task SemCom focused on full observation settings, our research explores a more realistic case where only distributed partial observations are available, such as in a production line monitored by multiple sensing nodes. To address this, we propose an SemCom system that supports multi-task processing through cooperation on the transmitter side via split structure and collaboration on the receiver side. We have used an information-theoretic perspective with variational approximations for our end-to-end data-driven approach. Simulation results demonstrate that the proposed co-operative and collaborative multi-task (CCMT) SemCom system significantly improves task execution accuracy, particularly in complex datasets, if the noise introduced from the communication channel is not limiting the task performance too much. Our findings contribute to a SemCom framework capable of handling distributed sources and multiple tasks simultaneously, advancing the applicability of SemCom systems in real-world scenarios.

In wireless communication systems, accurate Channel State Information (CSI) is essential for base stations to perform downlink precoding. While much of the existing research primarily focus on compressing the CSI matrix, they often neglect the impact of subsequent pre-transmission processes such as quantization, channel coding, and modulation. This paper investigates two distinct approaches for CSI dimensionality reduction: TransNet, a transformer-based neural network, and Dynamic Mode Decomposition (DMD), a mathematical decomposition technique for dynamical systems. We analyze how quantization, channel coding, and modulation affect CSI feedback for both methods. Unlike TransNet, DMD can decompose the channel matrix into components (called modes) with varying significance. This decomposition allows for an effective application of Unequal Error Protection (UEP) techniques to DMD modes, which is not feasible with TransNet-based CSI. Simulation results reveal that while the compression performance of TransNet and DMD varies based on factors like target CSI size and channel estimation error, integrating UEP techniques for DMD-based CSI yields superior CSI transmission performance compared to TransNet-based CSI.

In this study, we introduce an innovative signal to interference and noise ratio (SINR) time sequence feedback scheme based on vector quantization variational autoencoder (VQ-VAE). We compress the SINR sequence at the user equipment (UE) side and reconstruct it at the base station (BS) side. The reconstructed sequence is then utilized for SINR prediction at the BS. The VQ-VAE framework compresses SINR sequences into a compact embedding space involving several embedding vectors. Instead of transmitting the entire compressed SINR sequence back, we only need to transmit the index of the corresponding embedded vector. Based on the index, the sequence will be reconstructed. Moreover, a principal component analysis (PCA) based method is employed to reshape the distribution of the embedding space and compression performance is improved consequently. Our numerical simulations demonstrate that VQ-VAE combined with PCA achieves superior reconstruction and prediction accuracy while requiring fewer quantization bits compared to 3GPP commonly used method, differential quantization. Therefore, the proposed scheme is a promising solution for enhancing SINR sequence compression and prediction in wireless communication systems.

Low Earth Orbit (LEO) satellite-to-handheld connections herald a new era in satellite communications. Space-Division Multiple Access (SDMA) precoding is a method that mitigates interference among satellite beams, boosting spectral efficiency. While optimal SDMA precoding solutions have been proposed for ideal channel knowledge in various scenarios, addressing robust precoding with imperfect channel information has primarily been limited to simplified models. However, these models might not capture the complexity of LEO satellite applications. We use the Soft Actor-Critic (SAC) deep Reinforcement Learning (RL) method to learn robust precoding strategies without the need for explicit insights into the system conditions and imperfections. Our results show flexibility to adapt to arbitrary system configurations while performing strongly in terms of achievable rate and robustness to disruptive influences compared to analytical benchmark precoders.

This paper introduces an adaptive pilot design for wireless communication systems, dynamically adjusting the number of pilot symbols based on historical channel information. Focusing on Orthogonal Frequency Division Multiplexing (OFDM) systems formally adopted in 5G networks and is a potential candidate for future 6G systems, we employ Generative Adversarial Networks (GAN) to model and predict channel states by leveraging spectral and temporal correlations within the OFDM resource grid. The predicted channel state is then used alongside the available pilots to perform channel estimation. As the accuracy of channel predictions improves, fewer pilots are needed to achieve the desired channel estimation accuracy. Thus, we propose a scheme to estimate the GAN prediction accuracy, which is essential for determining the required number of pilots. Additionally, the proposed method supports ultra-low latency communication by performing channel estimation within the duration of a single OFDM symbol, making it highly suitable for beyond 5G and 6G networks. Simulation results demonstrate significant improvements in channel estimation accuracy compared to traditional fixed-pilot schemes, thereby reducing the number of required pilots to achieve a target channel accuracy.

The Nyquist-Shannon sampling theorem states that bandlimited signals can be perfectly reconstructed from samples taken at a fixed rate. Signals with varying spectral content are not considered, which leads to an unnecessarily high number of samples in signal intervals with narrowband content. An extension of the Nyquist-Shannon theorem enables the definition of variable bandwidth signals through nonlinear time axis distortion. This technique, known as time warping, enables variable-rate sampling based on instantaneous bandwidth, resulting in sample numbers proportional to the average bandwidth rather than the maximum bandwidth as in classical sampling. In practice, however, the instantaneous bandwidth of a signal is unknown, except for a few analytically determinable exceptions. In this paper, we introduce a novel spectrogram-based algorithm for estimating the instantaneous bandwidth of classically sampled signals, allowing to project them to variable bandwidth signals. We examine the tradeoff between sample reduction and reconstruction accuracy of electrocardiograms and compare the results to classical downsampling.

In multiple-antenna communication systems, it is crucial for the base station to acquire accurate downlink Channel State Information (CSI) to optimize signal transmission through beamforming. However, with the absence of the channel reciprocity, the mobile station must follow the process of channel estimation with feeding the CSI back to the base station. This can introduce a substantial overhead that increases with the number of antennas and the bandwidth. Therefore the CSI must be first compressed and quantized before reporting. In this paper we introduce a novel approach that based on combining Dynamic Mode Decomposition (DMD) with Residual Vector Quantization (RVQ). RVQ adapts the quantization accuracy based on the DMD output, namely the modes. This strategy allows the system to prioritize important feedback data and reduce the overhead bits needed for less critical data. Simulation results show that our approach can reduce the CSI feedback overhead while maintaining the target channel reconstruction accuracy.

“Network of subnetworks” is envisioned to be a key enabler in a 6G network with extremely low (100 μs) latency and high-reliability (99.9999%-99.99999%) in demanding applications. However, to achieve this level of performance, it is necessary to introduce a proactive and robust interference estimation considering the random mobility of subnetworks in an ultra-dense environment. We propose long-short term memory (LSTM) to learn the non-linear behavior of interference power time series for robust estimation and prediction in in-X subnetworks. This proposed method empowers to prediction/estimation of interference on the subnetwork itself. The achieved estimation result is compared with the moving average-based and expectation based estimators. Furthermore, we introduce federated learning (FL) in-X subnetworks' interference estimation, which learns cooperatively from the interference power vector of subnetworks participating in training. The results indicate that the proposed FL-based estimator achieves a higher convergence speed and lower estimation error.

In wireless communication, it is essential for the base station (BS) to obtain the downlink channel state information (CSI). In case of the absence of channel reciprocity, the mobile station (MS) needs to report the CSI back to the BS. In mobile multiple input multiple output (MIMO) systems, the CSI feedback overhead grows proportionally with the number of antennas and with the employed bandwidth. Moreover, the channel characteristics change constantly, so the feedback must be reported repeatedly with cautiously designed update intervals depending on how rapidly the channel changes. The increasing CSI overhead becomes a performance bottleneck, therefore it is vital to reduce it while keeping the system performance as good as required. In this paper, we propose a novel method based on designing a dynamical model of a time-varying channel with help of a framework called dynamic mode decomposition (DMD). Reporting the model to the BS gives it the ability to predict the channel state and track its changes over time. Simulation results show that the proposed method can increase the interval duration between the successive feedback updates and thus reduce the average overhead.

Modern wireless systems require strict key performance indicators (KPIs), such as very high reliability and throughput. Link adaptation (LA) is a core technology used to achieve these targets, with outer loop link adaptation (OLLA) being the most commonly used algorithm due to its feasibility and simplicity. OLLA uses a term called backoff factor to correct the signal to interference and noise ratio (SINR) estimate mapped from channel state information (CSI) to obtain an effective SINR. Based on the effective SINR, modulation and coding scheme (MCS) will be selected. OLLA adjusts the backoff factor with respect to hybrid automatic repeat request (HARQ) in a linear manner. However, this leads to effective SINR fluctuation and quite often results in overestimation. Hence, the performance of the system will be degraded. In this work, we propose a novel algorithm which introduces an adaptive adjustment step size for the backoff factor using an exponentially decaying factor to alleviate this issue. Since the backoff factor is not adjusted in a linear manner like OLLA, we call the proposed novel algorithm non-linear outer loop link adaptation (NOLLA). NOLLA can be regarded as an extension of OLLA that retains low complexity and high feasibility, providing the possibility to improve the link transmission in an uncomplicated way. Numerical evaluations demonstrate that NOLLA achieves higher reliability and throughput in scenarios with and without interference.

Wireless sensor nodes need a drastically reduced technical complexity to fit constraints of future applications. Reducing complexity often results in a degradation of energy and bandwidth efficiency. An interesting new approach that promises to reduce both technical complexity and energy consumption is event-based communication (EBC). While practical low-complexity implementations of such systems have already been proposed, the general question of energy and bandwidth efficiency remains open. In this paper, we compare these between EBC and a system relying on classical uniform sampling. We show that EBC is indeed much more energy efficient, and this comes at the cost of bandwidth efficiency. Therefore EBC is particularly suitable in combination with ultra-wideband communication.

Advances in mobile communication capabilities open the door for closer integration of pre-hospital and in-hospital care processes. For example, medical specialists can be enabled to guide on-site paramedics and can, in turn, be supplied with live vitals or visuals. Consolidating such performance-critical applications with the highly complex workings of mobile communications requires solutions both reliable and efficient, yet easy to integrate with existing systems. This paper explores the application of Deep Deterministic Policy Gradient (DDPG) methods for learning a communications resource scheduling algorithm with special regards to priority users. Unlike the popular Deep-Q-Network methods, the DDPG is able to produce continuous-valued output. With light post-processing, the resulting scheduler is able to achieve high performance on a flexible sum-utility goal.

Greater capabilities of mobile communications technology enable the interconnection of on-site medical care at a scale previously unavailable. However, embedding such critical, demanding tasks into the already complex infrastructure of mobile communications has proven challenging. This paper explores a resource allocation scenario where a scheduler must balance mixed performance metrics among connected users. To fulfill this resource allocation task, we present a scheduler that adaptively switches between different model-based scheduling algorithms. We make use of a deep Q-Network (DQN) to learn the benefit of selecting a scheduling paradigm for a given situation, combining advantages from model-driven and data-driven approaches. The resulting ensemble scheduler is able to combine its constituent algorithms to maximize a sum-utility cost function while ensuring performance on designated high-priority users.

Following the great success of Machine Learning (ML), especially Deep Neural Networks (DNNs), in many research domains in 2010s, several ML-based approaches were proposed for detection in large inverse linear problems, e.g., massive MIMO systems. The main motivation behind is that the complexity of Maximum A-Posteriori (MAP) detection grows exponentially with system dimensions. Instead of using DNNs, essentially being a black-box, we take a slightly different approach and introduce a probabilistic Continuous relaxation of disCrete variables to MAP detection. Enabling close approximation and continuous optimization, we derive an iterative detection algorithm: Concrete MAP Detection (CMD). Furthermore, extending CMD by the idea of deep unfolding into CMDNet, we allow for (online) optimization of a small number of parameters to different working points while limiting complexity. In contrast to recent DNN-based approaches, we select the optimization criterion and output of CMDNet based on information theory and are thus able to learn approximate probabilities of the individual optimal detector. This is crucial for soft decoding in today’s communication systems. Numerical simulation results in MIMO systems reveal CMDNet to feature a promising accuracy complexity trade-off compared to State of the Art. Notably, we demonstrate CMDNet’s soft outputs to be reliable for decoders.

Wireless industrial radio communication systems spawn a new set of requirements with focus on high reliability and low latency. These requirements were identified in the Industry 4.0 (I4.0) initiative as well as in 5th Generation (5G) mobile communication standardization in the form of Ultra Reliable Low Latency Communication (URLLC).Specifically Closed-Loop-Control (CLC) applications exhibit periodic deterministic communication with short packets. These applications require ultra low latency which bars the application of retransmissions to improve reliability. Also, many CLC applications are very sensitive to burst errors but can tolerate single packet loss. Therefore, we propose to shift the focus from sum-rate maximization to burst error minimization. As a first step, we perform an extensive burst error analysis of state of the art scheduling and Resource Allocation (RA) strategies. We show that any dynamic RA outperforms a static RA by a large margin.

The sampling of neural signals plays an important role in modern neuroscience, especially for prosthetics. However, due to hardware and data rate constraints, only spike trains can get recovered reliably. State of the art prosthetics can still achieve impressive results, but to get higher resolutions the used data rate needs to be reduced. In this paper, this is done by expressing the data with exponential and sinusoidal splines. As these signals have a finite number of degrees of freedom per unit of time, they can be analyzed and reconstructed with the Finite Rate of Innovation (FRI) framework. We show, that we can reduce the needed data rate by 90% to achieve the same resolution as without compression. Additionally, we propose analytic boundaries for the reconstruction of these splines and present an algorithm that guarantees the reconstruction within these boundaries. Furthermore, we test the algorithm on real neural stimuli.

Compressive Sensing Multi-User Detection (CS-MUD) is a recently developed physical (PHY) layer technique [1] to support Massive Machine Communication (MMC) in the next generation of mobile communication (5G) [2], [3]. CS-MUD has been investigated in joint Medium Access Control (MAC) and PHY layer protocol design [4], [5] but the lack of analytical performance description makes the joint design cumbersome. To expand the joint MAC- and PHY-layer protocol design considering a larger parameter space and to gain insights in the cross-layer optimization, we present a performance abstraction for CS-MUD to avoid extensive numerical simulation. Within this work, we exploit a low-complexity approach to approximate the performance of CS-MUD through the Replica Symmetry in large system analysis [6].

Massive Machine Type Communication is seen as one major driver for the research of new physical layer technologies for future communication systems. To handle massive access, the main challenges are avoiding control signaling overhead, low complexity data processing per sensor, supporting of diverse but rather low data rates and a flexible and scalable access. To address all these challenges, we propose a combination of compressed sensing based detection known as Compressed Sensing based Multi User Detection (CS-MUD) with multicarrier access schemes. We name this novel combination Multicarrier CS-MUD (MCSM). Previous investigations on CS-MUD facilitates massive direct random access by exploiting the signal sparsity caused by sporadic sensor activity. The new combined scheme MCSM with its flexibility in accessing time frequency resources additionally allows for either reducing the number of subcarriers or shortening the multicarrier symbol duration, i.e., we gain a high spectral efficiency. Simulation results are given to show the performance of the proposed scheme.

In slotted random access of many nodes, multi-user detection (MUD) can be applied to handle collisions. One novel PHY layer approach for jointly detecting activity and data in such a setting is Compressed Sensing based Multi-User Detection (CS-MUD). In this paper, we first summarize previous investigations on CS-MUD and subsequently propose two novel solutions for problems which have not yet been fully addressed: Firstly, we improve on previous results, by introducing a new approach which incorporates the channel decoder into the Compressed Sensing (CS) detector. Secondly, we analyze the resource efficiency of CS-MUD by adapting phase diagrams known from CS literature to the application of sporadic communication.

The emergence of Machine-to-Machine (M2M) communication requires new Medium Access Control (MAC) schemes and physical (PHY) layer concepts to support a massive number of access requests. The concept of coded random access, introduced recently, greatly outperforms other random access methods and is inherently capable to take advantage of the capture effect from the PHY layer. Furthermore, at the PHY layer, compressive sensing based multi-user detection (CS-MUD) is a novel technique that exploits sparsity in multi-user detection to achieve a joint activity and data detection. In this paper, we combine coded random access with CS-MUD on the PHY layer and show very promising results for the resulting protocol.

In this paper, a novel setpoint-based design approach for Irregular Repeat Accumulate (IRA) codes in iterative detection and decoding structures is presented. In contrast to conventional IRA code design in which the convolutional decoder is combined with the detector, the goal behind this approach is to keep the IRA decoding structure consisting of convolutional decoder and repetition decoder intact, i.e. to consider it as an inner loop of the overall detection structure. The outer loop is then composed of the IRA decoder and the system specific detector. This approach requires to adapt the irregular repetition code jointly to the convolutional decoder as well as to the detector which is achieved by formulating setpoints for the inner and outer code characteristic. As will be shown, the presented code design approach, although starting from a completely different viewpoint as the conventional approach, leads to an irregular repetition code with a very similar transfer characteristic and code rate than the conventional approach.

Performing joint activity and data detection has recently gained attention for reducing signaling overhead in multi-user Machine-to-Machine Communication systems. In this context, Compressed Sensing has been identified as a good candidate for joint activity and data detection especially in scenarios where the activity probability is very low. This paper augments activity and data detection for frame based multi-user uplink scenarios where nodes are (in)active for the duration of a frame. We propose a two stage detector which first estimates the set of active nodes followed by a data detector. Our detector outperforms symbol-by-symbol Maximum a posteriori detection.

The application of Compresses Sensing is a promising physical layer technology for the joint activity and data detection of signals. Detecting the activity pattern correctly has severe impact on the system performance and is therefore of major concern. In contrast to previous work, in this paper we optimize joint activity and data detection in under-determined systems by minimizing the Bayes-Risk for erroneous activity detection. We formulate a new Compressed Sensing Bayes-Risk detector which directly allows to influence error rates at the activity detection dynamically by a parameter that can be controlled at higher layers. We derive the detector for a general linear system and show that our detector outperforms classical Compressed Sensing approaches by investigating an overloaded CDMA system.

The growing field of Machine-to-Machine communication requires new physical layer concepts to meet future requirements. In previous works it has been shown for a synchronous CDMA transmission that Compressive Sensing (CS) detectors are capable of jointly detecting both activity and data in multi-user detection (MUD). However, many practical applications show some degree of asynchronicity. In order to reduce transmitter complexity, we propose an enhanced CS MUD that detects the delay in addition to activity and data. This solves synchronicity issues for scenarios with a known maximum delay, without requiring signaling or pre-compensation of asynchronicity.

Machine-to-Machine communication requires new physical layer concepts to meet future requirements. In previous works it has already been shown that Compressive Sensing (CS) detectors are capable of jointly detecting both activity and data in multi-user detection (MUD). For this detection we propose a new generalized Group Orthogonal Matching Pursuit algorithm that allows the use of additional side information regarding the sparsity structure. As a specific example, we exploit the information of a sparsity-aware Viterbi decoder in an iterative feedback loop to improve the activity detection. Here, a significant improvement of the activity detection is already achieved by executing only a single additional detection and decoding step.

The standardization of the fourth generation of mobile communication systems was mainly driven by the demands for higher data-rates and improved Quality of Service. To reach these goals interference coordination has been identified as a promising research field for better exploitation of the time and frequency resources. This paradigm shift from interference avoidance to interference coordination is also reflected in the ongoing enhancement of the 4th generation of mobile communication systems such as 3GPP Long Term Evolution. In this context, numerous investigations have focused on the allocation of precoding matrices that are part of the link adaptation process by some form of base station (eNB) coordination. Within this work we develop a non-centralized interference coordination scheme by noticing that the re-allocation of a precoding matrix can lead to an uncontrolled change of the interference level at users located in neighboring cells, especially at the edge. To this end, we provide a fully closed form mathematical framework describing these changes. Based on this, we derive a simple metric that improves the precoding matrix selection process in the User Equipment with the result that interference changes can be reduced without having any standard impact. This novel scheme can also be seen as an extension to previous inter-cell interference coordination schemes without the need of base-station cooperation.

Adaptive coding and modulation is an important topic considering future communication systems. Orthogonal frequency division multiplexing (ODFM) has been identified as a promising technique, which offers the possibility for further enhancements by bit and power loading schemes. Commonly, channel coding has not been considered in the optimization of such algorithms. It is, however, an important component used in nearly every communication system. In this paper we propose a new scheme to adapt code rate, modulation and transmit power by solving a convex optimization problem based on a bisection approach in order to enhance the frame error rate at a fixed target rate.