Two-Timescale End-to-End Learning for Channel Acquisition and Hybrid Precoding

Hu, Qiyu; Kang, Kai; Yu, Guanding; Hoydis, Jakob; Eldar, Yonina C.

doi:10.1109/jsac.2021.3126050

Cited by 21 publications

(16 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A joint hybrid precoder design procedure has been described in [18] for full-duplex relay-aided multiuser mmWave MIMO systems, considering also the effects of imperfect CSI. The authors of [19] and [20] successfully developed two-timescale hybrid precoding schemes for maximizing the sum-rate, and reducing both the complexity as well as CSI feedback overhead. A frame-based transmission scenario is considered in their work, wherein each frame comprises a fixed number of time slots.…”

Section: A Related Work and Contributionsmentioning

confidence: 99%

“…By contrast, the short-timescale baseband precoders are optimized for each time slot based on the low-dimensional effective CSI. Hence, an optimization based solution is developed in [19], whereas a deep neural network (DNN)-aided technique is designed in [20]. The angular-sparsity is also a key feature of the THz MIMO channel [5], [21], which arises due to the highly directional beams of large antenna arrays, coupled with high propagation losses and signal blockage in the THz regime.…”

Section: A Related Work and Contributionsmentioning

confidence: 99%

See 1 more Smart Citation

Hybrid Transceiver Design for Tera-Hertz MIMO Systems Relying on Bayesian Learning Aided Sparse Channel Estimation

Srivastava

Tripathi

Varshney

et al. 2023

IEEE Trans. Wireless Commun.

View full text Add to dashboard Cite

Hybrid transceiver design in multiple-input multiple-output (MIMO) Tera-Hertz (THz) systems relying on sparse channel state information (CSI) estimation techniques is conceived. To begin with, a practical MIMO channel model is developed for the THz band that incorporates its molecular absorption and reflection losses, as well as its non-line-ofsight (NLoS) rays associated with its diffused components. Subsequently, a novel CSI estimation model is derived by exploiting the angular-sparsity of the THz MIMO channel. This is followed by designing a sophisticated Bayesian learning (BL)-based approach for efficient estimation of the sparse THz MIMO channel. The Bayesian Cramer-Rao Lower Bound (BCRLB) is also determined for benchmarking the performance of the CSI estimation techniques developed. Finally, an optimal hybrid transmit precoder and receiver combiner pair is designed, which directly relies on the beamspace domain CSI estimates and only requires limited feedback. Finally, simulation results are provided for quantifying the improved mean square error (MSE), spectral-efficiency (SE) and bit-error rate (BER) performance for transmission on practical THz MIMO channel obtained from the HIgh resolution TRANsmission (HITRAN)-database.

show abstract

Section: A Related Work and Contributionsmentioning

confidence: 99%

Section: A Related Work and Contributionsmentioning

confidence: 99%

Hybrid Transceiver Design for Tera-Hertz MIMO Systems Relying on Bayesian Learning Aided Sparse Channel Estimation

Srivastava

Tripathi

Varshney

et al. 2023

IEEE Trans. Wireless Commun.

View full text Add to dashboard Cite

show abstract

“…The gradient of the SSCA algorithm is the same as the first term in (32) except that the hybrid beamformers are acquired by the HBDUN. The second term is the gradient of the NN which only exists in the deep-unfolding NN but not in the SSCA algorithm.…”

Section: B the Ssca Algorithm Induced Deep-unfolding Nnmentioning

confidence: 99%

“…Compared to traditional algorithms, deep learning-based techniques have much lower computational complexity and often do not require CSI. In [29]- [32], the authors designed hybrid beamforming by employing convolutional neural networks (CNNs) and multi-layer perception (MLP) which are referred to as black-box neural networks (NNs). However, these NNs have poor interpretability and many samples are required for training.…”

Section: Introductionmentioning

confidence: 99%

Mixed-Timescale Deep-Unfolding for Joint Channel Estimation and Hybrid Beamforming

Kang

et al. 2022

IEEE J. Select. Areas Commun.

Self Cite

View full text Add to dashboard Cite

In massive multiple-input multiple-output (MIMO) systems, hybrid analog-digital beamforming is an essential technique for exploiting the potential array gain without using a dedicated radio frequency chain for each antenna. However, due to the large number of antennas, the conventional channel estimation and hybrid beamforming algorithms generally require high computational complexity and signaling overhead. In this work, we propose an end-to-end deep-unfolding neural network (NN) joint channel estimation and hybrid beamforming (JCEHB) algorithm to maximize the system sum rate in time-division duplex (TDD) massive MIMO. Specifically, the recursive least-squares (RLS) algorithm and stochastic successive convex approximation (SSCA) algorithm are unfolded for channel estimation and hybrid beamforming, respectively. In order to reduce the signaling overhead, we consider a mixed-timescale hybrid beamforming scheme, where the analog beamforming matrices are optimized based on the channel state information (CSI) statistics offline, while the digital beamforming matrices are designed at each time slot based on the estimated low-dimensional equivalent CSI matrices. We jointly train the analog beamformers together with the trainable parameters of the RLS and SSCA induced deep-unfolding NNs based on the CSI statistics offline. During data transmission, we estimate the low-dimensional equivalent CSI by the RLS induced deep-unfolding NN and update the digital beamformers. In addition, we propose a mixed-timescale deep-unfolding NN where the analog beamformers are optimized online, and extend the framework to frequency-division duplex (FDD) systems where channel feedback is considered. Simulation results show that the proposed algorithm can significantly outperform conventional algorithms with reduced computational complexity and signaling overhead.

show abstract

“…Recently, the deep learning have attracted great attention and have been widely employed in wireless communications due to their satisfactory performance and low complexity [1]. The deep neural networks (DNNs) have been applied to symbol detection [2], beamforming [3], channel feedback [4], [5], and channel estimation [6]- [8]. In particular, channel correlation are captured by DNNs to improve the accuracy of channel estimation [6]- [8].…”

Section: Introduction a Prior Workmentioning

confidence: 99%

DDPG-Driven Deep-Unfolding with Adaptive Depth for Channel Estimation with Sparse Bayesian Learning

Hu,

Shi,

Cai

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Deep-unfolding neural networks (NNs) have received great attention since they achieve satisfactory performance with relatively low complexity. Typically, these deepunfolding NNs are restricted to a fixed-depth for all inputs. However, the optimal number of layers required for convergence changes with different inputs. In this paper, we first develop a framework of deep deterministic policy gradient (DDPG)-driven deep-unfolding with adaptive depth for different inputs, where the trainable parameters of deep-unfolding NN are learned by DDPG, rather than updated by the stochastic gradient descent algorithm directly. Specifically, the optimization variables, trainable parameters, and architecture of deep-unfolding NN are designed as the state, action, and state transition of DDPG, respectively. Then, this framework is employed to deal with the channel estimation problem in massive multiple-input multipleoutput systems. Specifically, first of all we formulate the channel estimation problem with an off-grid basis and develop a sparse Bayesian learning (SBL)-based algorithm to solve it. Secondly, the SBL-based algorithm is unfolded into a layer-wise structure with a set of introduced trainable parameters. Thirdly, the proposed DDPG-driven deep-unfolding framework is employed to solve this channel estimation problem based on the unfolded structure of the SBL-based algorithm. To realize adaptive depth, we design the halting score to indicate when to stop, which is a function of the channel reconstruction error. Furthermore, the proposed framework is extended to realize the adaptive depth of the general deep neural networks (DNNs). Simulation results show that the proposed algorithm outperforms the conventional optimization algorithms and DNNs with fixed depth with much reduced number of layers.

show abstract

Two-Timescale End-to-End Learning for Channel Acquisition and Hybrid Precoding

Cited by 21 publications

References 39 publications

Hybrid Transceiver Design for Tera-Hertz MIMO Systems Relying on Bayesian Learning Aided Sparse Channel Estimation

Hybrid Transceiver Design for Tera-Hertz MIMO Systems Relying on Bayesian Learning Aided Sparse Channel Estimation

Mixed-Timescale Deep-Unfolding for Joint Channel Estimation and Hybrid Beamforming

DDPG-Driven Deep-Unfolding with Adaptive Depth for Channel Estimation with Sparse Bayesian Learning

Contact Info

Product

Resources

About