Xu Shi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xu Shi has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Jintao Wang
Tsinghua University
19 shared publications - Jian Song
Tsinghua University
11 shared publications - Xuehan Wang
Tsinghua University
6 shared publications - Hengyu Zhang
Tsinghua University
4 shared publications - Haohan Wang
Tsinghua University
3 shared publications - Yashuai Cao
School of Intelligence Science and Technology, University of Science and Technology Beijing, Beijing, China
3 shared publications
Latest Publications
- 14 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. PP (00189316) , 1-8
Diffusion Bridge-Based Joint Estimation and Detection Design for Superimposed-Pilot OFDM Receiver
Superimposed-pilot (SP) transmission improves spectral efficiency but poses significant challenges for receiver design due to pilot–data interference, which conventional algorithms struggle to address effectively under fast-fading conditions. To address this issue, this paper proposes a diffusion bridge–based receiver for joint channel estimation and data detection. By formulating the received signals as one endpoint distribution and the transmitted data together with the channel impulse response as the other endpoint distribution, the conditional probability path between them is learned while implicitly exploiting OFDM structural constraints. To enable more stable and efficient training, the network is designed to directly predict the clean output, and the corresponding flow-velocity loss function for data detection is derived. Furthermore, a two-stage factor scheduler is introduced to balance the training objectives of channel estimation and data detection. Simulation results show that the proposed method achieves lower channel estimation mean square error (MSE) and data detection bit error rate (BER) than existing schemes, while exhibits sufficient generalization ability at higher subcarriers and modulation orders, demonstrating its effectiveness as a data-driven solution for reliable communication in dynamic environments.
- 14 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. PP (00189316) , 1-8
Diffusion Bridge-Based Joint Estimation and Detection Design for Superimposed-Pilot OFDM Receiver
Superimposed-pilot (SP) transmission improves spectral efficiency but poses significant challenges for receiver design due to pilot–data interference, which conventional algorithms struggle to address effectively under fast-fading conditions. To address this issue, this paper proposes a diffusion bridge–based receiver for joint channel estimation and data detection. By formulating the received signals as one endpoint distribution and the transmitted data together with the channel impulse response as the other endpoint distribution, the conditional probability path between them is learned while implicitly exploiting OFDM structural constraints. To enable more stable and efficient training, the network is designed to directly predict the clean output, and the corresponding flow-velocity loss function for data detection is derived. Furthermore, a two-stage factor scheduler is introduced to balance the training objectives of channel estimation and data detection. Simulation results show that the proposed method achieves lower channel estimation mean square error (MSE) and data detection bit error rate (BER) than existing schemes, while exhibits sufficient generalization ability at higher subcarriers and modulation orders, demonstrating its effectiveness as a data-driven solution for reliable communication in dynamic environments.
- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
Doppler Shift Keying Modulation for Uplink Multiple Access Over Doubly-Dispersive Channels
The delay- Doppler (DD) domain modulation has been regarded as one of the most competitive candidates to support wireless communications for emerging high-mobility applications in the sixth-generation mobile networks. Unfortunately, most of the existing designs for DD domain modulation suffer from high peak-to-average power ratio (PAPR) and unbearable detection complexity under uplink transmission since large time duration and bandwidth are required to guarantee high DD resolutions. To address these issues, the Doppler shift keying (DSK) modulation based on the orthogonal delay Doppler division multiplexing modulator is proposed in this paper, where the input-output characterization in the DD domain is fully exploited. The principle of the DSK transceiver is first established with the one-hot mapper and low-complexity iterative successive interference cancellation-maximum ratio combining detector for point-to-point scenarios. The proposed scheme is then generalized to the zero auto-correlation sequence-based implementation, which benefits the extension of multi-user (MU) uplink DSK frameworks. For uplink DSK transmission, Zadoff-Chu (ZC) sequences are adopted as the basis sequences. We optimize the assignment of ZC roots to different user equipments (UEs) by minimizing the maximum inter-user interference. This optimization process, which analyzes the root allocation, directly assigns a specific ZC sequence to each UE. The PAPR and bit error rate performance of the proposed DSK modulation with the low-complexity detector is finally verified by extensive simulation results under doubly-dispersive channels, which demonstrates the superiority of DSK modulation especially for uplink multiple access over doubly dispersive channels.
- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
Doppler Shift Keying Modulation for Uplink Multiple Access Over Doubly-Dispersive Channels
The delay- Doppler (DD) domain modulation has been regarded as one of the most competitive candidates to support wireless communications for emerging high-mobility applications in the sixth-generation mobile networks. Unfortunately, most of the existing designs for DD domain modulation suffer from high peak-to-average power ratio (PAPR) and unbearable detection complexity under uplink transmission since large time duration and bandwidth are required to guarantee high DD resolutions. To address these issues, the Doppler shift keying (DSK) modulation based on the orthogonal delay Doppler division multiplexing modulator is proposed in this paper, where the input-output characterization in the DD domain is fully exploited. The principle of the DSK transceiver is first established with the one-hot mapper and low-complexity iterative successive interference cancellation-maximum ratio combining detector for point-to-point scenarios. The proposed scheme is then generalized to the zero auto-correlation sequence-based implementation, which benefits the extension of multi-user (MU) uplink DSK frameworks. For uplink DSK transmission, Zadoff-Chu (ZC) sequences are adopted as the basis sequences. We optimize the assignment of ZC roots to different user equipments (UEs) by minimizing the maximum inter-user interference. This optimization process, which analyzes the root allocation, directly assigns a specific ZC sequence to each UE. The PAPR and bit error rate performance of the proposed DSK modulation with the low-complexity detector is finally verified by extensive simulation results under doubly-dispersive channels, which demonstrates the superiority of DSK modulation especially for uplink multiple access over doubly dispersive channels.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25, 21261-21275
BeamCKM: A Framework of Channel Knowledge Map Construction for Multi-Antenna Systems
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25, 21261-21275
BeamCKM: A Framework of Channel Knowledge Map Construction for Multi-Antenna Systems
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 4105-4109
Channel Knowledge Map-Aided CSI Feedback With PnP Module for Massive MIMO Systems
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 15, 4105-4109
Channel Knowledge Map-Aided CSI Feedback With PnP Module for Massive MIMO Systems
- 29 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 5446-5462
Meta-Hierarchical Reinforcement Learning-Based Beamforming for Near-Field Multi-User Communications
Extremely Large Antenna Array Systems (ELAAs) promise to provide ultra-high data rates for multi-user 6G Communications, leveraging massive antenna elements to enhance spatial diversity and spectral efficiency. Unfortunately, considering the propagation of spherical waves in ELAAs, near-field beamforming suffers from substantial performance degradation, diminished spatial focusing precision, considerable computational overhead, and increased system costs. To address these challenges, this paper develops an intelligent hybrid beamforming scheme to empower near-field multi-user THz communications with meta-hierarchical reinforcement learning (MHRL). On the one hand, a novel codebook is designed with Hierarchical Deep Q Network (H-DQN), dramatically releasing the potential of near-field beamforming in ELAAs. To explain the availability of our codebook, the quantization bound is analyzed along with the corresponding computational complexity. On the other hand, we further develop a meta-beamforming framework by integrating our intelligent codebook to meta-reinforcement learning (MRL)-based hybrid beamformer, accelerating the CSI acquisition for multi-user communication. For deep investigation, the rate loss of our scheme is calculated against the ideal continuous beamformer, which reveals that the quantization error has been properly bounded. Simulation results show that the proposed algorithm can efficiently manage the near-field beam with higher beamforming gain, and the proposed beamforming framework exhibits improved spectral efficiency, as well as increased robustness to ELAAs.
- 29 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 5446-5462
Meta-Hierarchical Reinforcement Learning-Based Beamforming for Near-Field Multi-User Communications
Extremely Large Antenna Array Systems (ELAAs) promise to provide ultra-high data rates for multi-user 6G Communications, leveraging massive antenna elements to enhance spatial diversity and spectral efficiency. Unfortunately, considering the propagation of spherical waves in ELAAs, near-field beamforming suffers from substantial performance degradation, diminished spatial focusing precision, considerable computational overhead, and increased system costs. To address these challenges, this paper develops an intelligent hybrid beamforming scheme to empower near-field multi-user THz communications with meta-hierarchical reinforcement learning (MHRL). On the one hand, a novel codebook is designed with Hierarchical Deep Q Network (H-DQN), dramatically releasing the potential of near-field beamforming in ELAAs. To explain the availability of our codebook, the quantization bound is analyzed along with the corresponding computational complexity. On the other hand, we further develop a meta-beamforming framework by integrating our intelligent codebook to meta-reinforcement learning (MRL)-based hybrid beamformer, accelerating the CSI acquisition for multi-user communication. For deep investigation, the rate loss of our scheme is calculated against the ideal continuous beamformer, which reveals that the quantization error has been properly bounded. Simulation results show that the proposed algorithm can efficiently manage the near-field beam with higher beamforming gain, and the proposed beamforming framework exhibits improved spectral efficiency, as well as increased robustness to ELAAs.
- 24 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (11) , 3400-3404
Energy Efficiency Optimization for Symbol-Level Precoding-Aided RSMA MIMO System
Rate-splitting multiple access (RSMA) decomposes transmitted signals into common and private streams, employing successive interference cancellation (SIC) to cancel interference from common to private streams, thereby improving robustness under diverse channel conditions. Conventional RSMA implementations based on linear precoding do not actively employ residual interference, while symbol-level precoding (SLP) and certain interference-aware techniques provide one promising pathway for further interference exploitation. This work presents three key contributions: (1) we introduce a two-stage interference coordination framework, which first eliminates common signal interference through SIC, and then utilizes the constructive interference; (2) we integrate the SLP design with interference exploitation to formulate an energy efficiency maximization problem; (3) we develop an optimization algorithm for the non-convex problem in RSMA-SLP systems, which decomposes the original formulation into tractable problems via fractional programming for efficient iterative solutions. Our simulation confirms that the proposed scheme not only enhances system robustness but also improves energy efficiency through interference exploitation and verifies that the proposed scheme outperforms existing multiple-access precoding methods.
- 24 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications Letters
- Vol. 14 (11) , 3400-3404
Energy Efficiency Optimization for Symbol-Level Precoding-Aided RSMA MIMO System
Rate-splitting multiple access (RSMA) decomposes transmitted signals into common and private streams, employing successive interference cancellation (SIC) to cancel interference from common to private streams, thereby improving robustness under diverse channel conditions. Conventional RSMA implementations based on linear precoding do not actively employ residual interference, while symbol-level precoding (SLP) and certain interference-aware techniques provide one promising pathway for further interference exploitation. This work presents three key contributions: (1) we introduce a two-stage interference coordination framework, which first eliminates common signal interference through SIC, and then utilizes the constructive interference; (2) we integrate the SLP design with interference exploitation to formulate an energy efficiency maximization problem; (3) we develop an optimization algorithm for the non-convex problem in RSMA-SLP systems, which decomposes the original formulation into tractable problems via fractional programming for efficient iterative solutions. Our simulation confirms that the proposed scheme not only enhances system robustness but also improves energy efficiency through interference exploitation and verifies that the proposed scheme outperforms existing multiple-access precoding methods.
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24 (7) , 6180-6194
Orthogonal Hyperbolic Frequency Division Multiplexing Modulation for Underwater Acoustic Communications
Though meaningful progress has been achieved for mobile wireless networks with the development of orthogonal frequency division multiplexing (OFDM) modulation, the reliability and transmission efficiency of underwater acoustic (UWA) communications are still limited, which cannot support the ever-increasing requirements of underwater applications. The major barrier lies in the wideband time-varying channels with extremely large time scales. Since the time dilation or contraction of baseband signals cannot be ignored in UWA communications, the orthogonality between subcarriers of OFDM is destroyed more significantly than in narrowband high-mobility channels, which leads to severe performance degradation. To solve this problem, a novel multicarrier modulation scheme referred to as the orthogonal hyperbolic frequency division multiplexing (OHFDM) modulation is proposed in this paper inspired by the scale-invariance of hyperbolic frequency signals, where a series of orthogonal narrowband hyperbolic frequency subcarriers (HFSs) is adopted to load data symbols. The input-output relation is then characterized by jointly processing the carrier and subcarrier signals, and selecting the appropriate sampling time of the output of matched filters at the receiver corresponding to the time scale of the wideband time-varying channel. The analysis reveals that the approximate orthogonality can be guaranteed, i.e., much smaller inter-carrier-interference (ICI) than OFDM systems, which enhances the system reliability and reduces the processing complexity at the receiver notably. The robustness of the proposed OHFDM modulation when path-specific scales are involved is also confirmed theoretically in this paper. Simulation results demonstrate that the proposed OHFDM modulation outperforms OFDM in terms of bit error rate (BER) under typical UWA channels with large time scales.
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24 (7) , 6180-6194
Orthogonal Hyperbolic Frequency Division Multiplexing Modulation for Underwater Acoustic Communications
Though meaningful progress has been achieved for mobile wireless networks with the development of orthogonal frequency division multiplexing (OFDM) modulation, the reliability and transmission efficiency of underwater acoustic (UWA) communications are still limited, which cannot support the ever-increasing requirements of underwater applications. The major barrier lies in the wideband time-varying channels with extremely large time scales. Since the time dilation or contraction of baseband signals cannot be ignored in UWA communications, the orthogonality between subcarriers of OFDM is destroyed more significantly than in narrowband high-mobility channels, which leads to severe performance degradation. To solve this problem, a novel multicarrier modulation scheme referred to as the orthogonal hyperbolic frequency division multiplexing (OHFDM) modulation is proposed in this paper inspired by the scale-invariance of hyperbolic frequency signals, where a series of orthogonal narrowband hyperbolic frequency subcarriers (HFSs) is adopted to load data symbols. The input-output relation is then characterized by jointly processing the carrier and subcarrier signals, and selecting the appropriate sampling time of the output of matched filters at the receiver corresponding to the time scale of the wideband time-varying channel. The analysis reveals that the approximate orthogonality can be guaranteed, i.e., much smaller inter-carrier-interference (ICI) than OFDM systems, which enhances the system reliability and reduces the processing complexity at the receiver notably. The robustness of the proposed OHFDM modulation when path-specific scales are involved is also confirmed theoretically in this paper. Simulation results demonstrate that the proposed OHFDM modulation outperforms OFDM in terms of bit error rate (BER) under typical UWA channels with large time scales.
- 5 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (5) , 3441-3455
Double-Sided Near-Field XL-MIMO: Beamfocusing Codeword Selection and Channel Estimation
In the double-sided near-field extremely large-scale multi-input multi-output (XL-MIMO) systems, due to the spherical-wavefront propagation, the line-of-sight (LoS) path exhibits multiple independent propagation components, leading to a channel rank greater than one. In contrast, the non-line-of-sight (NLOS) path is typically dominated by a single propagation component. Consequently, the unified modeling of mixed LoS and NLoS paths remains unresolved, particularly when with non-parallel and non-coplanar uniform linear arrays (ULAs) at the transceivers. Furthermore, there exist bottlenecks in the beamforming codeword design and low-overhead estimation in double-sided near-field communications. In this paper, we present a unified channel model to characterize both LoS and NLoS paths in extremely large-scale multiple-input multiple-output (XL-MIMO) systems. Apart from the transmitter (Tx)-side and receiver (Rx)-side separate response vectors, an additional Tx/Rx-coupled term with a Vandermonde windowing pattern is studied for the XL-MIMO LoS path. Two codebook-based beamfocusing schemes are proposed, which are termed as the beamspace-projection and diagonal-decomposition schemes. The achievable spectral efficiency and average power are theoretically analyzed in closed form. Under this framework, we further propose a low-overhead unified LoS/NLoS orthogonal matching pursuit (UOMP) algorithm for XL-MIMO channel estimation, which is then extended via 3-stage multiple-measurement-vector (3S-MMV) for complexity reduction. Last, simulation results demonstrate the superiority of the proposed strategies in both beamforming codeword selection and sparse channel estimation.
- 5 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (5) , 3441-3455
Double-Sided Near-Field XL-MIMO: Beamfocusing Codeword Selection and Channel Estimation
In the double-sided near-field extremely large-scale multi-input multi-output (XL-MIMO) systems, due to the spherical-wavefront propagation, the line-of-sight (LoS) path exhibits multiple independent propagation components, leading to a channel rank greater than one. In contrast, the non-line-of-sight (NLOS) path is typically dominated by a single propagation component. Consequently, the unified modeling of mixed LoS and NLoS paths remains unresolved, particularly when with non-parallel and non-coplanar uniform linear arrays (ULAs) at the transceivers. Furthermore, there exist bottlenecks in the beamforming codeword design and low-overhead estimation in double-sided near-field communications. In this paper, we present a unified channel model to characterize both LoS and NLoS paths in extremely large-scale multiple-input multiple-output (XL-MIMO) systems. Apart from the transmitter (Tx)-side and receiver (Rx)-side separate response vectors, an additional Tx/Rx-coupled term with a Vandermonde windowing pattern is studied for the XL-MIMO LoS path. Two codebook-based beamfocusing schemes are proposed, which are termed as the beamspace-projection and diagonal-decomposition schemes. The achievable spectral efficiency and average power are theoretically analyzed in closed form. Under this framework, we further propose a low-overhead unified LoS/NLoS orthogonal matching pursuit (UOMP) algorithm for XL-MIMO channel estimation, which is then extended via 3-stage multiple-measurement-vector (3S-MMV) for complexity reduction. Last, simulation results demonstrate the superiority of the proposed strategies in both beamforming codeword selection and sparse channel estimation.
- 18 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 23 (10) , 14335-14349
Beamforming Design for Massive MIMO-Aided Over-the-Air Computation: A Mutual Information Perspective
Over-the-air computation (AirComp) is considered a transformative enabler for next-generation artificial intelligence (AI) services and wireless data aggregation via the electromagnetic waveform-superposition property of wireless multi-access channels (MAC). However, the conventional distortion metric, minimum square error (MSE), is imperfect and not universally applicable in specific AirComp scenarios in a low-signal-to-noise ratio (SNR) regime and under power budget constraint. Conversely, the average discriminant gain is studied for task-oriented AirComp AI services like classification but with inaccurate performance indication. To solve these problems, this work establishes a novel framework for AirComp systems from the mutual information (MI) perspective. First, we categorize the AirComp model into two distinct classes based on the source (sensing) data independence, namely diverse-targets (DT) AirComp and homogeneous-target (HT) AirComp . Both categories with different inputs like classical Gaussian and classification-based Gaussian mixture model (GMM), can be unified and assessed via MI criterion. Next, for the DT AirComp system, we introduce a novel MI-aided AirComp beamforming scheme employing majorization-minimization (MM) relaxation. As for the HT AirComp, we present a heuristic successive approximation (SA)-based beamforming method considering complex GMM inputs. We also provide the feedback and update protocol for AirComp tracking. Simulations validate the superior performance on AirComp throughput and task-oriented metrics such as classification accuracy with our proposed MI-aided beamforming schemes.
- 18 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 23 (10) , 14335-14349
Beamforming Design for Massive MIMO-Aided Over-the-Air Computation: A Mutual Information Perspective
Over-the-air computation (AirComp) is considered a transformative enabler for next-generation artificial intelligence (AI) services and wireless data aggregation via the electromagnetic waveform-superposition property of wireless multi-access channels (MAC). However, the conventional distortion metric, minimum square error (MSE), is imperfect and not universally applicable in specific AirComp scenarios in a low-signal-to-noise ratio (SNR) regime and under power budget constraint. Conversely, the average discriminant gain is studied for task-oriented AirComp AI services like classification but with inaccurate performance indication. To solve these problems, this work establishes a novel framework for AirComp systems from the mutual information (MI) perspective. First, we categorize the AirComp model into two distinct classes based on the source (sensing) data independence, namely diverse-targets (DT) AirComp and homogeneous-target (HT) AirComp . Both categories with different inputs like classical Gaussian and classification-based Gaussian mixture model (GMM), can be unified and assessed via MI criterion. Next, for the DT AirComp system, we introduce a novel MI-aided AirComp beamforming scheme employing majorization-minimization (MM) relaxation. As for the HT AirComp, we present a heuristic successive approximation (SA)-based beamforming method considering complex GMM inputs. We also provide the feedback and update protocol for AirComp tracking. Simulations validate the superior performance on AirComp throughput and task-oriented metrics such as classification accuracy with our proposed MI-aided beamforming schemes.
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (11) , 19852-19865
Frequency-Scanning-Based Fast Multiuser Beam Training for Wideband Massive MIMO
Beam squint effect causes severe performance degradation for wideband beamforming inside Internet of Things (IoT) communication, and the true-time-delay (TTD) line has been regarded as a promising enabler to address this issue. However, beam training becomes a challenging puzzle in TTD-aided transceivers, where enormous beam directions bring about unacceptable training overhead, especially for the overhead-efficient IoT devices. In this article, based on a joint delay-phase beamforming structure, we provide enhanced frequency-scanning-based training schemes for remarkable overhead reduction. Simultaneous beams pointing to different physical directions over a set of OFDM subcarriers can be generated to reduce overhead. The pointing directions can be flexibly controlled following two subcarrier-angular mapping policies: 1) forward pairing and 2) backward pairing. Besides, the power leakage problem is retrieved via the compressive phase retrieval (CPR) method to avoid beam mismatch. Furthermore, we adopt it into the multiuser scenario and propose a frequency-scanning-based simultaneous multiuser beam training (FS-MBT) scheme. The different subcarriers’ pencil beams illuminate a broad angular sector, while several sectors are merged into multifinger wide beams for simultaneous multiuser training. Analytical and numerical results demonstrate the proposed schemes’ superiority over existing methods in both single-user and multiuser scenarios.
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (11) , 19852-19865
Frequency-Scanning-Based Fast Multiuser Beam Training for Wideband Massive MIMO
Beam squint effect causes severe performance degradation for wideband beamforming inside Internet of Things (IoT) communication, and the true-time-delay (TTD) line has been regarded as a promising enabler to address this issue. However, beam training becomes a challenging puzzle in TTD-aided transceivers, where enormous beam directions bring about unacceptable training overhead, especially for the overhead-efficient IoT devices. In this article, based on a joint delay-phase beamforming structure, we provide enhanced frequency-scanning-based training schemes for remarkable overhead reduction. Simultaneous beams pointing to different physical directions over a set of OFDM subcarriers can be generated to reduce overhead. The pointing directions can be flexibly controlled following two subcarrier-angular mapping policies: 1) forward pairing and 2) backward pairing. Besides, the power leakage problem is retrieved via the compressive phase retrieval (CPR) method to avoid beam mismatch. Furthermore, we adopt it into the multiuser scenario and propose a frequency-scanning-based simultaneous multiuser beam training (FS-MBT) scheme. The different subcarriers’ pencil beams illuminate a broad angular sector, while several sectors are merged into multifinger wide beams for simultaneous multiuser training. Analytical and numerical results demonstrate the proposed schemes’ superiority over existing methods in both single-user and multiuser scenarios.