Artificial intelligence (AI) has emerged as a promising tool for channel state information (CSI) feedback. While recent research primarily focuses on improving feedback accuracy on a specific dataset through novel architectures, the underlying mechanism of AI-based CSI feedback remains unclear. This study explores the mechanism through analyzing performance across diverse datasets, with findings suggesting that superior feedback performance stems from AI models’ strong fitting capabilities and their ability to leverage environmental knowledge. Building on these findings, we propose a prompt-enabled large AI model (LAM) for CSI feedback. The LAM employs powerful Transformer blocks and is trained on extensive datasets from various scenarios. Meanwhile, the channel distribution (environmental knowledge), represented as the mean of channel magnitude in the angular-delay domain, is incorporated as a scenario-specific prompt within the decoder to further enhance reconstruction quality. Simulation results confirm that the proposed prompt-enabled LAM significantly improves feedback accuracy and generalization performance while reducing data collection requirements in new scenarios.

Multi-user millimeter-wave communication relies on narrow beams and dense cell deployments to ensure reliable connectivity. However, tracking optimal beams for multiple mobile users across multiple base stations (BSs) results in significant signaling overhead. Recent works have explored the capability of out-of-band (OOB) modalities in obtaining spatial characteristics of wireless channels and reducing pilot overhead in single-BS single-user/multi-user systems. However, applying OOB modalities for multi-BS selection towards dense cell deployments leads to high coordination overhead, i.e, excessive computing overhead and high latency in data exchange. How to leverage OOB modalities to eliminate pilot overhead and achieve efficient multi-BS coordination in multi-BS systems remains largely unexplored. In this paper, we propose a novel OOB modality synergy (OMS) based mobility management scheme to realize multi-user beam prediction and proactive BS selection by synergizing two OOB modalities, i.e., vision and location. Specifically, mobile users are initially identified via spatial alignment of visual sensing and location feedback, and then tracked according to the temporal correlation in image sequence. Subsequently, a binary encoding map based gain and beam prediction network (BEM-GBPN) is designed to predict beamforming gains and optimal beams for mobile users at each BS, such that a central unit can control the BSs to perform user handoff and beam switching. Simulation results indicate that the proposed OMS-based mobility management scheme enhances beam prediction and BS selection accuracy and enables users to achieve 91% transmission rates of the optimal with zero pilot overhead and significantly improve multi-BS coordination efficiency compared to existing methods.

Channel State Information (CSI) provides a detailed description of the wireless channel and has been widely adopted for Wi-Fi sensing, particularly for high-precision indoor positioning. However, complete CSI is rarely available in real-world deployments due to hardware constraints and the high communication overhead required for feedback. Moreover, existing positioning models lack mechanisms to detect when users move outside their trained regions, leading to unreliable estimates in dynamic environments. In this paper, we present FPNet, a unified deep learning framework that jointly addresses channel feedback compression, accurate indoor positioning, and robust anomaly detection (AD). FPNet leverages the beamforming feedback matrix (BFM), a compressed CSI representation natively supported by IEEE 802.11ac/ax/be protocols, to minimize feedback overhead while preserving critical positioning features. To enhance reliability, we integrate ADBlock, a lightweight AD module trained on normal BFM samples, which identifies out-of-distribution scenarios when users exit predefined spatial regions. Experimental results using standard 2.4 GHz Wi-Fi hardware show that FPNet achieves positioning accuracy above 97% with only 100 feedback bits, boosts net throughput by up to 22.92%, and attains AD accuracy over 99% with a false alarm rate below 1.5%. These results demonstrate FPNet’s ability to deliver efficient, accurate, and reliable indoor positioning on commodity Wi-Fi devices.

Reconfigurable Intelligent Surfaces (RIS) are an emerging technology that holds significant promise for customizing wireless channels to meet specific communication requirements. Accurate channel state information (CSI) is essential for fully realizing the potential of RIS. However, due to the passive nature of RIS and the large number of reflecting elements, acquiring CSI for the base station (BS)-RIS-user equipment (UE) link presents considerable challenges. In this paper, we propose a deep learning (DL)-based framework for downlink CSI acquisition. Specifically, we introduce a novel DL-based channel estimation framework, termed PPNet, which facilitates efficient pilot transmission. The key innovation of PPNet lies in the joint design and optimization of pilot signals from the BS and phase shifts from the RIS, both represented through neural networks, alongside the channel estimation module. By capturing environment-specific features with neural networks, PPNet enables more efficient utilization of pilot power. Furthermore, we propose a beamforming-oriented CSI acquisition framework, RIS-E2ENet, which jointly optimizes the entire CSI acquisition process, including channel estimation, CSI feedback, and active/passive beamforming design, to enhance CSI acquisition efficiency. To adapt to the dynamic nature of real-world environments, RIS-E2ENet incorporates a lightweight UE-side neural network design, enabling low-overhead online training. Extensive evaluations show that the proposed frameworks improve spectral efficiency by 58.55%, while maintaining minimal pilot and feedback overhead.

To reduce channel acquisition overhead, spatial, time, and frequency-domain channel extrapolation techniques have been widely studied. In this paper, we propose a novel deep learning-based Position-domain Channel Extrapolation framework (named PCEnet) for cell-free massive multiple-input multiple-output (MIMO) systems. The user’s position, which contains significant channel characteristic information, can greatly enhance the efficiency of channel acquisition. In cell-free massive MIMO, while the propagation environments between different base stations and a specific user vary and their respective channels are uncorrelated, the user’s position remains constant and unique across all channels. Building on this, the proposed PCEnet framework leverages the position as a bridge between channels to establish a mapping between the characteristics of different channels, thereby using one acquired channel to assist in the estimation and feedback of others. Specifically, this approach first utilizes neural networks (NNs) to infer the user’s position from the obtained channel. The estimated position, shared among BSs through a central processing unit (CPU), is then fed into an NN to design pilot symbols and concatenated with the feedback information to the channel reconstruction NN to reconstruct other channels, thereby significantly enhancing channel acquisition performance. Additionally, we propose a simplified strategy where only the estimated position is used in the reconstruction process without modifying the pilot design, thereby reducing latency. Furthermore, we introduce a position label-free approach that infers the relative user position instead of the absolute position, eliminating the need for ground truth position labels during the localization NN training. Simulation results demonstrate that the proposed PCEnet framework reduces pilot and feedback overheads by up to 50%.

The Channel Quality Indicator (CQI) is a fundamental component of channel state information (CSI) that enables adaptive modulation and coding by selecting the optimal modulation and coding scheme to meet a target block error rate. While AI-enabled CSI feedback has achieved significant advances, especially in precoding matrix index feedback, AI-based CQI feedback remains underexplored. Conventional subband-based CQI approaches, due to coarse granularity, often fail to capture fine frequency-selective variations and thus lead to suboptimal resource allocation. In this paper, we propose an AI-driven subcarrier-level CQI feedback framework tailored for 6G and NextG systems. First, we introduce CQInet, an autoencoder-based scheme that compresses per-subcarrier CQI at the user equipment and reconstructs it at the base station, significantly reducing feedback overhead without compromising CQI accuracy. Simulation results show that CQInet increases the effective data rate by 7.6% relative to traditional subband CQI under equivalent feedback overhead. Building on this, we develop SR-CQInet, which leverages super-resolution to infer fine-grained subcarrier CQI from sparsely reported CSI reference signals (CSI-RS). SR-CQInet reduces CSI-RS overhead to 3.5% of CQInet's requirements while maintaining comparable throughput. These results demonstrate that AI-driven subcarrier-level CQI feedback can substantially enhance spectral efficiency and reliability in future wireless networks.

Recent advancements in foundation models (FMs) have attracted increasing attention in the wireless communication domain. Leveraging the powerful multi-task learning capability, FMs hold the promise of unifying multiple tasks of wireless communication with a single framework. Nevertheless, existing wireless FMs face limitations in the uniformity to address multiple tasks with diverse inputs/outputs across different communication scenarios. In this paper, we propose a MUlti-taSk Environment-aware FM (MUSE-FM) with a unified architecture to handle multiple tasks in wireless communications, while effectively incorporating scenario information. Specifically, to achieve task uniformity, we propose a unified prompt-guided data encoder-decoder pair to handle data with heterogeneous formats and distributions across different tasks. Besides, we integrate the environmental context as a multi-modal input, which serves as prior knowledge of environment and channel distributions and facilitates cross-scenario feature extraction. Simulation results illustrate that the proposed MUSE-FM outperforms existing methods for various tasks, and its prompt-guided encoder-decoder pair facilitates few-shot adaptation to new task configurations. Moreover, the incorporation of environment information improves the ability to adapt to different scenarios.

Large language models (LLMs) have achieved remarkable success across a wide range of tasks, particularly in natural language processing and computer vision. This success naturally raises an intriguing yet unexplored question: Can LLMs be harnessed to tackle channel state information (CSI) compression and feedback in massive multiple-input multiple-output (MIMO) systems? Efficient CSI feedback is a critical challenge in next-generation wireless communication. In this paper, we pioneer the use of LLMs for CSI compression, introducing a novel framework that leverages the powerful denoising capabilities of LLMs—capable of error correction in language tasks—to enhance CSI reconstruction performance. To effectively adapt LLMs to CSI data, we design customized pre-processing, embedding, and post-processing modules tailored to the unique characteristics of wireless signals. Extensive numerical results demonstrate the promising potential of LLMs in CSI feedback, opening up possibilities for this research direction.

Reconfigurable Intelligent Surface (RIS) is a promising technology with the potential to tailor wireless channels to specific communication needs. In RIS-assisted communications, channel estimation has long been a challenge due to the passive nature and the large number of RIS elements. In this paper, we introduce a novel deep learning-based downlink channel estimation framework, named PPNet, which facilitates efficient pilot transmission. The core innovation of PPNet lies in the joint design and optimization of pilot signals from the base station and phase shifts from the RIS, both of which are represented using neural networks, together with the channel estimation module. Simulation results show that the proposed framework significantly improves the estimation performance with limited pilot overhead.

Accurate channel state information (CSI) is critical for realizing the full potential of multiple-antenna wireless communication systems. While deep learning (DL)-based CSI feedback methods have shown promise in reducing feedback overhead, their generalization capability across varying propagation environments remains limited due to their data-driven nature. Existing solutions based on online training improve adaptability but impose significant overhead in terms of data collection and computational resources. In this work, we propose AdapCsiNet, an environment-adaptive DL-based CSI feedback framework that eliminates the need for online training. By integrating environmental information—represented as a scene graph—into a hypernetwork-guided CSI reconstruction process, AdapCsiNet dynamically adapts to diverse channel conditions. A two-step training strategy is introduced to ensure baseline reconstruction performance and effective environment-aware adaptation. Simulation results demonstrate that AdapCsiNet achieves up to 46.4% improvement in CSI reconstruction accuracy and matches the performance of online learning methods without incurring additional runtime overhead.

Wi-Fi sensing has gained substantial attention in recent years, particularly for indoor positioning applications. Conventional indoor wireless positioning methods typically assume access to complete channel state information (CSI), which is often impractical in real-world systems. This paper proposes a novel approach for indoor positioning utilizing a compressed beamforming feedback matrix (BFM), which is inherently integrated into the Wi-Fi protocol for meeting CSI feedback requirements. We introduce FPNet, a joint neural network model, in which the BFM is compressed into codewords by an encoder at the client station (STA) side. These codewords are subsequently transmitted to the access point (AP) side, which features a decoder and a positioning network responsible for reconstructing the codewords and estimating positions. The encoder and decoder are trained end-to-end. FPNet is implemented with standard Wi-Fi equipment operating in the 2.4 GHz band. Experimental results demonstrate that this approach not only improves net throughput by up to 22.92% but also achieves positioning accuracy exceeding 97%.

In the domain of reconfigurable intelligent surface (RIS)-assisted wireless communications, efficient channel state information (CSI) feedback is crucial. This paper proposes RIS-CoCsiNet, a novel deep learning-based framework aimed at significantly enhancing feedback efficiency. The proposed method leverages the inherent correlation among neighboring user equipments (UEs) by categorizing RIS-UE CSI information into two parts: shared information among nearby UEs and unique information specific to each individual UE. By exploiting the correlation in RIS-UE CSI, redundant transmission of shared information can be substantially reduced, thereby minimizing the overhead associated with repeatedly feeding back this shared data. Unlike conventional autoencoder-based CSI feedback frameworks, our approach incorporates an additional decoder and a combination neural network (NN) at the base station. These components recover the shared information from the feedback CSI of two neighboring UEs and combine it with the individual information, respectively, without requiring any modifications at the UEs. Through end-to-end learning, the encoders at neighboring UEs are trained to collaboratively feedback shared information while independently feeding back the unique information. For UEs equipped with multiple antennas, a baseline NN architecture with long short-term memory (LSTM) modules is introduced to capture the correlation among nearby antennas. Additionally, since the RIS-UE CSI phase is not sparse, we propose magnitude-dependent phase feedback strategies that incorporate statistical or instantaneous CSI magnitude information into the phase feedback process. Extensive simulations across two diverse channel datasets validate the effectiveness of RIS-CoCsiNet.

Artificial intelligence (AI)-based channel state information (CSI) prediction, aimed at enhancing CSI accuracy and reducing overhead, has shown significant advancements over traditional model-based prediction methods. Despite these advantages, its practical deployment has been hindered by unreliable prediction performance due to AI instability. This study introduces a reliable AI-based CSI prediction framework by implementing a proxy-based performance monitoring mechanism. Specifically, we deploy a lightweight proxy at the user equipment (UE), trained via knowledge distillation to accommodate the UE’s limited capacities. This proxy mimics the output of the CSI prediction network at the base station (BS) side, enabling the UE to monitor the accuracy of the predicted CSI and prevent undesirable outcomes. To overcome the deployment challenges in operational systems, we detail the practical implementation procedures of our proposed method, covering both offline training and online operation phases. Simulation results show that our proxy-based monitor can achieve over 90% consistency with the CSI prediction network at the BS side and avoid over 85% of unsatisfactory prediction outcomes under various practical considerations, demonstrating remarkable generalization capabilities across different configurations.

Artificial intelligence (AI) has emerged as a promising tool in channel state information (CSI) feedback tasks. Although current research primarily focuses on improving feedback accuracy through innovative AI approaches, the reliability of these systems in real-world scenarios often goes overlooked. Specifically, a closer examination of the feedback accuracy of individual CSI samples reveals significant variations, underscoring the imperative need for performance monitoring of AI-based CSI feedback. Building upon this observation, we introduce a pragmatic framework for AI-based CSI feedback. This process involves assessing feedback accuracy (i.e., conducting performance monitoring) on the user side before transmitting the CSI codeword. In particular, this method utilizes a lightweight proxy decoder, trained via knowledge distillation, to emulate the mapping function of the original decoder at the base station. The goal is to generate, at the user end, CSI identical to that produced at the base station by the original, more powerful decoder, thus enable precise prediction of feedback accuracy. Simulation results demonstrate that our proposed performance monitoring method can precisely predict feedback accuracy with low complexity and accurately detect low-quality feedback samples with a detection rate of nearly 95%, ensuring reliable transmission.

Ultra-reliable low-latency communication is the key technology for smart factories and autonomous vehicles. However, traditional beam training approaches in millimeter-wave communications generally cause significant latency and communication overhead, especially in the case of multi-user communications. To tackle this problem, we propose a novel Vision-aided Multi-user Beam Tracking (VA-MUBT) framework for mmWave massive MIMO system, which leverages deep learning based visual object detection and multiple objects tracking algorithm to enable fast beam tracking of multi-user. In addition, a prototype is constructed to evaluate the proposed VA-MUBT framework and the experimental results based on this prototype show that the accuracy of 3-time beam search can reach near 90% with only 8% overhead of the exhaustive beam search method. Hence, the proposed VA-MUBT demonstrates the superiority in achieving fast multi-user beam tracking and significantly reducing the communication overhead.

Traditional indoor location data sharing methods using centralized servers face issues like safe and reliable transmission, personal privacy leaks, location information tampering, and computing and storage loads, hampering the growth of personalized indoor services. In this paper, a novel mobile blockchain-enabled federated learning (MBFL) data sharing framework for indoor positioning is presented. Then, we derive training latency and reward of the individual user, and formulate latency-limited resource allocation as a non-cooperative game. We propose an efficient alternating iterative algorithm to achieve the Nash equilibrium of this game. Numerical results demon-strate that the proposed alternating iterative algorithm achieves rapid convergence. Furthermore, when confronted with model poisoning attacks, the MBFL method exhibits superior security performance compared to the traditional FL method.

Deep learning has shown promise in enhancing channel state information (CSI) feedback. However, many studies indicate that better feedback performance often accompanies higher computational complexity. Pursuing better performance-complexity tradeoffs is crucial to facilitate practical deployment, especially on computation-limited devices, which may have to use lightweight autoencoder with unfavorable performance. To achieve this goal, this paper introduces knowledge distillation (KD) to achieve better tradeoffs, where knowledge from a complicated teacher autoencoder is transferred to a lightweight student autoencoder for performance improvement. Specifically, two methods are proposed for implementation. Firstly, an autoencoder KD-based method is introduced by training a student autoencoder to mimic the reconstructed CSI of a pretrained teacher autoencoder. Secondly, an encoder KD-based method is proposed to reduce training overhead by performing KD only on the student encoder. Additionally, a variant of encoder KD is introduced to protect user equipment and base station vendor intellectual property. Numerical simulations demonstrate that the proposed methods can significantly improve the student autoencoder’s performance, while reducing the number of floating point operations and inference time to 3.05%–5.28% and 13.80%–14.76% of the teacher network, respectively. Furthermore, the variant encoder KD method effectively enhances the student autoencoder’s generalization capability across different scenarios, environments, and bandwidths.

Traditional indoor location information management methods based on centralized servers have problems such as safe and reliable transmission, personal privacy leaks, location information tampering, and computing and storage loads. These problems have seriously affected the development of personalized services based on indoor location information. In this paper, a novel mobile blockchain-enabled federated learning (MBFL) information management framework for indoor positioning is presented, comprising the mobile blockchain model, the federated learning (FL) model, and the InterPlanetary file storage model. Then, we design the MBFL algorithm, establishing a robust foundation for collaborative model training, efficient block mining, and secure data storage. Moreover, we derive training and mining latency as well as the individual user rewards, and formulate latency-limited resource allocation strategies as a non-cooperative game. We propose an efficient alternating iterative algorithm to achieve the Nash equilibrium of this game. Numerical results demonstrate that the proposed alternating iterative algorithm achieves rapid convergence and strikes an effective balance between economic and time efficiency. Furthermore, when confronted with model poisoning attacks, the MBFL algorithm exhibits superior security performance compared to the traditional FL algorithm. Future work will focus on adapting the MBFL framework for various indoor environments and enhancing consumption and computational efficiency with hybrid consensus mechanisms.

Deep learning (DL)-based channel state information (CSI) feedback has garnered significant research attention in recent years. However, previous research has overlooked the potential privacy disclosure problem caused by transmitting CSI datasets during the training process. In this study, we introduce a federated edge learning (FEEL)-based training framework for DL-based CSI feedback. This approach differs from the conventional centralized learning (CL)-based framework, where the CSI datasets are collected at the base station (BS) before training. Instead, each user equipment (UE) trains a local autoencoder network and exchanges model parameters with the BS. This approach provides better protection for data privacy compared to CL. To further reduce communication overhead in FEEL, we quantize the uplink and downlink model transmission into different bits based on their influence on feedback performance. Additionally, since the heterogeneity of CSI datasets among different UEs can degrade the performance of the FEEL-based framework, we introduce a personalization strategy to enhance feedback performance. This strategy allows for local fine-tuning to adapt the global model to the channel characteristics of each UE. Simulation results indicate that the proposed personalized FEEL-based training framework can significantly improve the performance of DL-based CSI feedback while reducing communication overhead.

Most learning-based channel state information (CSI) feedback efforts concentrate on enhancing feedback accuracy through innovative neural network (NN) designs and exploiting correlations. This paper introduces an integrated learning framework for CSI feedback and localization designed to synergistically improve both tasks. We present a novel unified approach for CSI feedback and downlink CSI-based localization, where feedback is facilitated by an autoencoder, and the downlink CSI-based localization uses the feedback codeword directly without requiring reconstruction. The goal is to simultaneously minimize feedback and localization errors. Additionally, for users with access to coarse position data, we propose a refined framework that integrates this information into both the feedback mechanism and localization processes. This coarse positional knowledge is incorporated into the encoding and decoding stages to reduce feedback errors and is inputted into the localization NN to enhance localization accuracy. The improved framework is refined through an end-to-end training strategy, focusing on concurrently reducing feedback and localization errors. Simulation results using ray tracing channel datasets demonstrate that our proposed method not only enables feedback and localization tasks to mutually benefit but also shows that incorporating coarse positional data significantly increases the accuracy of both CSI feedback and CSI-based localization.

Deep learning has been introduced to improve implicit channel state information (CSI) feedback, and it significantly outperforms codebook-based feedback methods used in existing systems. This study proposes a multi-domain correlation-aided implicit CSI feedback framework that uses deep learning. This framework retains the existing implicit feedback mechanism while introducing the aid of the multi-domain correlation property of CSI matrices to the feedback process for performance improvement. First, a time correlation-aided implicit feedback framework is proposed, where the correlation among adjacent CSI matrices is exploited to improve the CSI reconstruction accuracy. Second, to utilize the correlation between the uplink and downlink channel, the uplink channel magnitude is introduced into the CSI reconstruction process at the base station. Additionally, the framework combines the aid of time and bidirectional channel correlation properties to further enhance performance. Simulation results show that, with the aid of the multi-domain correlation property, the feedback overhead can be reduced by 75% and 85% compared to approaches without correlation utilization and Type II codebook, respectively.

Recent advances in artificial intelligence offer groundbreaking alternatives to conventional codebook-based channel state information (CSI) feedback techniques. Confronted with the influx of CSI data from simulations and real-world environments, leveraging neural networks to mine valuable insights poses significant training costs and technical challenges for base station (BS) manufacturers. To address this, we propose a third-party platform serving as a CSI knowledge repository and feedback model hub, reducing training expenses and addressing technical issues for various BS manufacturers. However, tailoring training for each manufacturer’s model may lead to proprietary information leaks and inefficient resource utilization. In response, we present “CSI Meta-knowledge Support”, a cutting-edge CSI feedback network deployment strategy using Learngene, enabling seamless transfer of CSI meta-knowledge across heterogeneous networks. This method captures a Learngene unit enriched with vital CSI meta-knowledge during comprehensive training sessions, serving as a plug-and-play prior to facilitate swift convergence and efficient local fine-tuning for manufacturers. The approach introduces adaptable and scalable CSI feedback network configurations, emphasizing reusability, cost-effectiveness, and resource management while safeguarding intellectual property. Our tests demonstrate enhanced performance, reduced training sample demands, and faster convergence relative to conventional techniques.

Deep learning has brought about a revolution in the design of the channel state information (CSI) feedback module in wireless communications. However, designing the optimal neural network (NN) architecture for CSI feedback can be a laborious and time-consuming process, and manual design can be prohibitively expensive for customized NNs tailored to different scenarios. To tackle this challenge, this paper proposes the use of neural architecture search (NAS) to automate the generation of scenario-customized CSI feedback NN architectures. By employing automated machine learning and gradient-descent-based NAS, an efficient and cost-effective architecture design process is achieved, requiring less expert experience and design time, thus lowering the design threshold. The proposed approach leverages implicit scene knowledge and integrates it into the scenario customization process in a data-driven manner, fully exploiting the potential of deep learning in a given scenario. To address the issue of excessive search, early stopping and elastic selection mechanisms are employed, further enhancing the proposed scheme. The experimental results demonstrate that the generated architecture, known as Auto-CsiNet, outperforms manually-designed models in terms of reconstruction performance (achieving approximately 14% improvement) and complexity (reducing by approximately 50%), highlighting the effectiveness of the NAS-based automatic scheme. Furthermore, the paper analyzes the impact of the scenario on the NN architecture and capacity.

Accurate and effective channel state information (CSI) feedback is a key technology for massive multiple-input and multiple-output systems. Recently, deep learning (DL) has been introduced for CSI feedback enhancement through massive collected training data and lengthy training time, which is quite costly and impractical for realistic deployment. In this article, a knowledge-driven meta-learning approach is proposed, where the DL model initialized by the meta model obtained from meta training phase is able to achieve rapid convergence when facing a new scenario during target retraining phase. Specifically, instead of training with massive data collected from various scenarios, the meta task environment is constructed based on the intrinsic knowledge of spatial-frequency characteristics of CSI for meta training. Moreover, the target task dataset is also augmented by exploiting the knowledge of statistical characteristics of wireless channel, so that the DL model can achieve higher performance with small actually collected dataset and short training time. In addition, we provide analyses of rationale for the improvement yielded by the knowledge in both phases. Simulation results demonstrate the superiority of the proposed approach from the perspective of feedback performance and convergence speed.

Deep learning has been introduced to implicit channel state information (CSI) feedback and considerably outperforms codebook-based feedback methods adopted by existing systems. This work proposes a time correlation-aided deep learning-based implicit CSI feedback framework named Tbi-ImCsiNet. The long short-term memory network is introduced into the implicit CSI compression side and reconstruction side to extract and utilize the time correlation property among CSI matrices and improve the framework performance. Simulation results show that the proposed Tbi-ImCsiNet reduces approximately 58.3% of the feedback overhead compared with the method without time correlation utilization.

Deep learning (DL) has shown great potential in channel state information (CSI) feedback and localization. In this paper, a DL-based integrated CSI feedback and localization framework called FLnet, in which the feedback and localization tasks complement each other, is proposed. Specifically, unlike the existing works that sequentially realize the above two tasks, FLnet jointly designs the autoencoder-based feedback and deep neural networks (DNN)-based localization tasks. The encoder at the user equipment (UE) compresses and quantizes the downlink CSI. Then, the decoder and the DNN at the base station reconstruct the downlink CSI and predict the location of the UE based on the feedback information, respectively. The feedback and localization modules are trained together by an end-to-end approach. Simulation results show that the localization error of FLnet is reduced by 30% compared with that of the separate design while the feedback performance is comparable or even improved.

Deep learning has revolutionized the design of channel state information (CSI) feedback modules in wireless communication. However, designing an optimal neural network (NN) architecture for CSI feedback can be laborious and time-consuming, especially for customized networks targeting different scenarios. To address this challenge, this paper proposes the use of Neural Architecture Search (NAS) to automatically generate scenario-specific CSI feedback neural network architectures. By employing automated machine learning and gradient-based NAS, an efficient and cost-effective architecture design process is achieved with reduced reliance on expert knowledge and design time, thus lowering the design threshold. This approach leverages implicit scenario knowledge and integrates it into the scenario customization process in a data-driven manner, fully harnessing the potential of deep learning in a given scenario. Experimental results demonstrate that the generated architecture called Auto-CsiNet outperforms manually designed models in terms of reconstruction performance (improvement by approximately 14%) and complexity reduction (approximately 50%), highlighting the effectiveness of NAS-based automated solutions.

For frequency division duplex (FDD) systems, downlink channel state information (CSI) feedback is essential. Deep learning-based auto-encoder (AE) structures have shown promise in reducing feedback overhead. However, designing a super-large AE network to handle the CSI of all scenarios is not practical. A more practical approach is to divide the CSI dataset by region/scenario and use multiple simple AE networks. However, this method requires high memory capacity, making it unsuitable for low-end user equipment (UE). In this paper, we propose a new UE-friendly framework based on multi-tasking mode. Our framework, called single-encoder-to-multiple-decoders (S-to-M), uses multi-task-learning to design multiple independent AEs into a joint architecture with a shared encoder that corresponds to multiple task-specific decoders. We also integrate GateNet as a classifier to enable the base station to autonomously select the right task-specific decoder for the subregion. Our experiments on a simulated multi-scenario CSI dataset show that our proposed S-to-M framework outperforms other benchmark modes by significantly reducing model complexity and UE memory consumption.

The downlink channel state information (CSI) feedback occupies substantial precious transmission resources in frequency-division duplexing (FDD) systems. In this work, we propose a data hiding-based CSI feedback framework, namely, EliCsiNet, to eliminate the CSI feedback overhead in FDD systems with deep learning. The key idea of this work is to hide downlink CSI within the transmitted messages (e.g., images) with no transmission resource occupation and few effects on the message semantic. We propose a novel neural network framework, in which the user extracts and hides the CSI features within the images by networks, and the base station recovers the CSI from the transmitted images. Simulation results demonstrate that the proposed EliCsiNet framework can eliminate the CSI feedback overhead with few effects on the transmitted images, including the image quality and classification accuracy.

Channel state information (CSI) plays a crucial role in the capacity of multiple-input and multiple-output systems, but CSI feedback occupies substantial precious transmission resources in frequency-division duplexing (FDD) systems. In this work, we propose a data hiding-based CSI feedback framework, namely, EliCsiNet, to eliminate the CSI feedback overhead in FDD systems through deep learning. The key idea is to hide/superimpose CSI in transmitted messages (e.g., images) with no transmission resource occupation and few effects on message semantics. Concretely, we introduce a novel neural network framework in which the user extracts and hides CSI features in images, and the base station recovers the CSI from the transmitted images. However, the essential source coding (e.g., JPEG compression) before data transmission causes two problems in the proposed EliCsiNet framework when applied in practical systems. First, the compression inevitably disturbs the information of the hidden CSI in images and affects the CSI reconstruction accuracy. Therefore, a two-stage separable training strategy, which includes coding-free end-to-end and coding-aware decoder-only training, is adopted to reduce these effects. Second, the bit length of the images coded via JPEG is unpredictable and uncontrollable, and CSI superimposition may lead to an increase in the bit length of the coded images. To avoid this issue, we divide a full image into several sub-blocks and select the one with the smallest length increment. Image entropy is also introduced to accelerate block selection. Simulation results demonstrate that the proposed EliCsiNet framework can eliminate the CSI feedback overhead with few effects on the features properties of transmitted images, including image quality and bit length.

Massive multiple-input multiple-output can obtain more performance gain by exploiting the downlink channel state information (CSI) at the base station (BS). Therefore, studying CSI feedback with limited communication resources in frequency-division duplexing systems is of great importance. Recently, deep learning (DL)-based CSI feedback has shown considerable potential. However, the existing DL-based explicit feedback schemes are difficult to deploy because current fifth-generation mobile communication protocols and systems are designed based on an implicit feedback mechanism. In this paper, we propose a DL-based implicit feedback architecture to inherit the low-overhead characteristic, which uses neural networks (NNs) to replace the precoding matrix indicator (PMI) encoding and decoding modules. By using environment information, the NNs can achieve a more refined mapping between the precoding matrix and the PMI compared with codebooks. The correlation between subbands is also used to further improve the feedback performance. Simulation results show that, for a single resource block (RB), the proposed architecture can save 25.0% – 40.0% of overhead compared with the Type I codebook under different antenna configurations. For a wideband system with 52 RBs, overhead can be saved by 30.7% and 48.0% compared with the Type II codebook when ignoring and considering extracting subband correlation, respectively.

The autoencoder empowered by artificial intelligence has shown considerable potential in solving channel state information (CSI) feedback problems in frequency-division duplexing systems. However, this method needs to completely change the existing feedback schemes, which is difficult to deploy in the next few years. This paper proposes an environment knowledge-aided codebook-based CSI feedback framework, which retains the existent codebook-based scheme while introducing environment knowledge to feedback process through neural networks (NNs) at the base station. Only an NN-based refining operation is added after the common standardized feedback approach. The NNs learn to automatically extract environment features and utilize the channel statistics through large volumes of recorded data. The NNs also use the partial correlation between bidirectional channels to further improve feedback performance. In addition, to deal with downlink channel estimation errors, we propose two strategies to reduce their effects using an NN-based denoise module. The proposed framework can be easily embedded in most existing codebook-based feedback methods, such as random vector quantization. Two channel datasets generated by QuaDRiGa and measured in practical systems are adopted to evaluate the proposed methods. Results show that the proposed method offers over 100% increase in the throughput compared with the baseline codebook because of more accurate feedback.

In frequency-division duplexing systems, the downlink channel state information (CSI) acquisition scheme leads to high training and feedback overhead. In this work, we propose an uplink-aided downlink channel acquisition framework using deep learning to reduce such overhead. We consider the entire downlink CSI acquisition process, including the downlink pilot design, channel estimation, and feedback. First, we propose an adaptive pilot design module by exploiting the correlation in magnitude among bidirectional channels in the angular domain to improve channel estimation. Second, to avoid the bit allocation problem during the feedback module, we concatenate the complex channel and embed the uplink channel magnitude to the channel reconstruction at the base station. Finally, we combine the two modules and compare two popular uplink-aided downlink channel acquisition frameworks. One framework estimates and subsequently feeds back the channel at the user equipment. In the other framework, the user equipment directly feeds back the received pilot signals to the base station. Results reveal that with the help of the uplink channel, directly feeding back pilot signals can save approximately 20% of feedback bits. This work thus provides a guideline for future research.

Many performance gains achieved by massive multiple-input and multiple-output depend on the accuracy of the downlink channel state information (CSI) at the transmitter (base station), which is usually obtained by estimating at the receiver (user equipment) and feeding back to the transmitter. The overhead of CSI feedback occupies substantial uplink bandwidth resources, especially when the number of transmit antennas is large. Deep learning (DL)-based CSI feedback refers to CSI compression and reconstruction by a DL-based autoencoder and can greatly reduce feedback overhead. In this paper, a comprehensive overview of state-of-the-art research on this topic is provided, beginning with basic DL concepts widely used in CSI feedback and then categorizing and describing some existing DL-based feedback works. The focus is on novel neural network architectures and utilization of communication expert knowledge to improve CSI feedback accuracy. Works on joint design of CSI feedback with other communication modules are also introduced, and some practical issues, including bitstream generation, multirate feedback, imperfect feedback, NN complexity, training dataset collection, online training, and standardization effect, are discussed. At the end of the paper, some challenges and potential research directions associated with DL-based CSI feedback in future wireless communication systems are identified.

In frequency-division duplexing systems, the downlink channel state information (CSI) should be fed back through an uplink transmission to reap the benefits of the massive multiple-input and multiple-output system, thereby leading to a large feedback overhead. The autoencoder-based architecture empowered by artificial intelligence has shown considerable potential in solving the CSI feedback problem. This method, however, needs to completely change the existing feedback schemes, which is difficult to deploy in the next few years. In this paper, we propose an environment knowledge-aided codebook-based CSI feedback framework, which retains the existent codebook-based scheme while introducing environment knowledge to the feedback process through neural networks (NNs) at the base station. The NNs learn to automatically extract the environment features and utilize the channel statistics through large volumes of recorded data. The channel dataset, which is generated by QuaDRiGa software, is adopted to evaluate the proposed methods. Results show that the proposed method offers over 100% increase in the throughput compared with the baseline feedback codebook because of the more accurate CSI feedback.

The potentials of massive multiple-input multiple-output (MIMO) are all based on the available instantaneous channel state information (CSI) at the base station (BS). Therefore, the user in frequency-division duplexing (FDD) systems has to keep on feeding back the CSI to the BS, thereby occupying large uplink transmission resources. Recently, deep learning (DL) has achieved great success in the CSI feedback. However, the existing works just focus on improving the feedback accuracy and ignore the effects on the following modules, e.g., beamforming (BF). In this paper, we propose a DL-based CSI feedback framework for BF design, called CsiFBnet. The key idea of the CsiFBnet is to maximize the BF performance gain rather than the feedback accuracy. We apply it to two representative scenarios: single- and multi-cell systems. The CsiFBnet-s in the single-cell system is based on the autoencoder architecture, where the encoder at the user compresses the CSI and the decoder at the BS generates the BF vector. The CsiFBnet-m in the multi-cell system has to feed back two kinds of CSI: the desired and the interfering CSI. The entire neural networks are trained by an unsupervised learning strategy. Simulation results show the great performance improvement and complexity reduction of the CsiFBnet compared with the conventional DL-based CSI feedback methods.

Millimeter-wave (mmWave) cloud radio access networks (CRANs) provide new opportunities for accurate cooperative localization, in which large bandwidths and antenna arrays and increased densities of base stations enhance the delay and angular resolution. This study considers the joint location and velocity estimation of user equipment (UE) and scatterers in a three-dimensional mmWave CRAN architecture. Several existing works have achieved satisfactory results by using neural networks (NNs) for localization. However, the black box NN localization method has limited robustness and accuracy and relies on a prohibitive amount of training data to increase localization accuracy. Thus, we propose a model-based learning network for localization to address these problems. In comparison with the black box NN, we combine NNs with geometric models. Specifically, we first develop an unbiased weighted least squares (WLS) estimator by utilizing hybrid delay and angular measurements, which determine the location and velocity of the UE in only one estimator, and can obtain the location and velocity of scatterers further. The proposed estimator can achieve the Cramér-Rao lower bound under small measurement noise and outperforms other state-of-the-art methods. Second, we establish a NN-assisted localization method called NN-WLS by replacing the linear approximations in the proposed WLS localization model with NNs to learn the higher-order error components, thereby enhancing the performance of the estimator, especially in a large noise environment. The solution possesses the powerful learning ability of the NN and the robustness of the proposed geometric model. Moreover, the ensemble learning is applied to improve the localization accuracy further. Comprehensive simulations show that the proposed NN-WLS is superior to the benchmark methods in terms of localization accuracy, robustness, and required time resources.

Massive multiple-input multiple-output (MIMO) is a promising technology to increase link capacity and energy efficiency. However, these benefits are based on available channel state information (CSI) at the base station (BS). Therefore, user equipment (UE) needs to keep on feeding CSI back to the BS, thereby consuming precious bandwidth resource. Large-scale antennas at the BS for massive MIMO seriously increase this overhead. In this paper, we propose a multiple-rate compressive sensing neural network framework to compress and quantize the CSI. This framework not only improves reconstruction accuracy but also decreases storage space at the UE, thus enhancing the system feasibility. Specifically, we establish two network design principles for CSI feedback, propose a new network architecture, CsiNet+, according to these principles, and develop a novel quantization framework and training strategy. Next, we further introduce two different variable-rate approaches, namely, SM-CsiNet+ and PM-CsiNet+, which decrease the parameter number at the UE by 38.0% and 46.7%, respectively. Experimental results show that CsiNet+ outperforms the state-of-the-art network by a margin but only slightly increases the parameter number. We also investigate the compression and reconstruction mechanism behind deep learning-based CSI feedback methods via parameter visualization, which provides a guideline for subsequent research.