Sheng Zhou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sheng Zhou has a total of 317 co-authors. Here are the top co-authors by number of shared publications:
- Zhisheng Niu
Tsinghua University
71 shared publications Zhisheng Niu
Beijing National Research Center for Information Science and Technology Department of Electronic Engineering Tsinghua University 100084 Beijing China
63 shared publications- Yuxuan Sun
Beijing Jiaotong University
25 shared publications - Zhiyuan Jiang
Shanghai Key Laboratory of Chips and Systems for Intelligent Connected Vehicle and the School of Communication and Information Engineering, Shanghai University, Shanghai, China
25 shared publications Jie Gong
Sun Yat-sen University
15 shared publications- Zhaojun Nan
Department of Electronic Engineering, Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing, China
12 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (11) , 19508-19522
Mobility-Aware Asynchronous Federated Learning With Dynamic Sparsification
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (11) , 19508-19522
Mobility-Aware Asynchronous Federated Learning With Dynamic Sparsification
- 12 September 2026
- Published by Springer Nature
- p. 206-224
DH-VLM: Dual-Horizon Cooperative Latent Reasoning for Autonomous Driving
- 12 September 2026
- Published by Springer Nature
- p. 206-224
DH-VLM: Dual-Horizon Cooperative Latent Reasoning for Autonomous Driving
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. PP (15361233) , 1-12
SLIDE: Simultaneous Model Downloading and Inference at the Wireless Network Edge
To support on-device inference, the next-generation mobile networks are expected to support real-time model downloading services to mobile users. However, powerful AI models typically have large model sizes, resulting in excessive end-to-end (E2E) downloading-and-inference (DAI) latency. To address this issue, we propose a simultaneous model downloading and inference (SLIDE) framework, which allows users to perform inference with downloaded layers while simultaneously receiving the remaining layers of the model. To this end, we formulate a task throughput maximization problem by jointly optimizing model provisioning, spectrum bandwidth allocation, and computing resource allocation for multi-user downlink systems. Unlike traditional DAI frameworks, SLIDE introduces recursive dependencies across layers, where inference latency depends recursively on the downloading bandwidth and computing resource allocation for each of the preceding layers. To solve this challenging problem, we design an efficient algorithm that acquires the optimal solution with polynomial-time complexity. Simulation results demonstrate that the proposed SLIDE framework significantly improves task throughput under latency and communication resource constraints compared with the conventional model downloading schemes.
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. PP (15361233) , 1-12
SLIDE: Simultaneous Model Downloading and Inference at the Wireless Network Edge
To support on-device inference, the next-generation mobile networks are expected to support real-time model downloading services to mobile users. However, powerful AI models typically have large model sizes, resulting in excessive end-to-end (E2E) downloading-and-inference (DAI) latency. To address this issue, we propose a simultaneous model downloading and inference (SLIDE) framework, which allows users to perform inference with downloaded layers while simultaneously receiving the remaining layers of the model. To this end, we formulate a task throughput maximization problem by jointly optimizing model provisioning, spectrum bandwidth allocation, and computing resource allocation for multi-user downlink systems. Unlike traditional DAI frameworks, SLIDE introduces recursive dependencies across layers, where inference latency depends recursively on the downloading bandwidth and computing resource allocation for each of the preceding layers. To solve this challenging problem, we design an efficient algorithm that acquires the optimal solution with polynomial-time complexity. Simulation results demonstrate that the proposed SLIDE framework significantly improves task throughput under latency and communication resource constraints compared with the conventional model downloading schemes.
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (6) , 2010-2014
Venusian: Rapid Wireless Baseband Validation via High-Level Programming and FPGA-Based RISC-V Accelerator Co-Design
Wireless communication systems demand increasingly agile and high-performance platforms to meet evolving protocol and deployment requirements. However, existing platforms often face tradeoffs between software (SW) flexibility for algorithm exploration and hardware (HW) efficiency for link-level evaluation. This brief presents Venusian, an agile HW/SW co-design testbed that bridges this gap for wireless baseband processing. Venusian features a lightweight Python-to-HW compilation framework that supports code reuse across simulation and deployment. On the HW side, it integrates a digital front-end (DFE) for real-time signal acquisition and a customized RISC-V-based vector processing unit (VPU) for accelerating rule-based baseband processing. The architecture is deployed on an evaluation board and validated under a 5-G new radio (NR) broadcast channel (BCH) decoding task. Experimental results demonstrate that Venusian achieves speedups of at least $3.61\times $ and $51.57\times $ over Intel and Arm platforms, respectively.
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (6) , 2010-2014
Venusian: Rapid Wireless Baseband Validation via High-Level Programming and FPGA-Based RISC-V Accelerator Co-Design
Wireless communication systems demand increasingly agile and high-performance platforms to meet evolving protocol and deployment requirements. However, existing platforms often face tradeoffs between software (SW) flexibility for algorithm exploration and hardware (HW) efficiency for link-level evaluation. This brief presents Venusian, an agile HW/SW co-design testbed that bridges this gap for wireless baseband processing. Venusian features a lightweight Python-to-HW compilation framework that supports code reuse across simulation and deployment. On the HW side, it integrates a digital front-end (DFE) for real-time signal acquisition and a customized RISC-V-based vector processing unit (VPU) for accelerating rule-based baseband processing. The architecture is deployed on an evaluation board and validated under a 5-G new radio (NR) broadcast channel (BCH) decoding task. Experimental results demonstrate that Venusian achieves speedups of at least $3.61\times $ and $51.57\times $ over Intel and Arm platforms, respectively.
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (8) , 13276-13290
FedCGD: Collective Gradient Divergence Optimized Scheduling for Wireless Federated Learning
Federated learning (FL) is a promising paradigm for multiple devices to cooperatively train a model. When applied in wireless networks, two issues consistently affect the performance of FL, i.e., data heterogeneity of devices and limited bandwidth. Many papers have investigated device scheduling strategies considering the two issues. However, most of them recognize data heterogeneity as a property of individual devices. In this paper, we prove that the convergence speed of FL is affected by the sum of device-level and sample-level collective gradient divergence (CGD). Device-level CGD refers to the gradient divergence of the scheduled device group, instead of the sum of the individual device divergence. Sample-level CGD is statistically upper bounded by sampling variance, which is inversely proportional to the total number of samples scheduled for local update. To derive a tractable form of the device-level CGD, we further consider classification tasks and transform it into the weighted earth moving distance (WEMD) between the group distribution and the global distribution. Then we propose FedCGD algorithm to minimize the sum of sampling variance and WEMD on classification tasks by device scheduling and bandwidth allocation, within polynomial time. Simulation shows that the proposed strategy increases classification accuracy on the CIFAR-10 dataset by up to 4.2% while scheduling 41.8% fewer devices, and flexibly switches between reducing WEMD and reducing sampling variance.
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (8) , 13276-13290
FedCGD: Collective Gradient Divergence Optimized Scheduling for Wireless Federated Learning
Federated learning (FL) is a promising paradigm for multiple devices to cooperatively train a model. When applied in wireless networks, two issues consistently affect the performance of FL, i.e., data heterogeneity of devices and limited bandwidth. Many papers have investigated device scheduling strategies considering the two issues. However, most of them recognize data heterogeneity as a property of individual devices. In this paper, we prove that the convergence speed of FL is affected by the sum of device-level and sample-level collective gradient divergence (CGD). Device-level CGD refers to the gradient divergence of the scheduled device group, instead of the sum of the individual device divergence. Sample-level CGD is statistically upper bounded by sampling variance, which is inversely proportional to the total number of samples scheduled for local update. To derive a tractable form of the device-level CGD, we further consider classification tasks and transform it into the weighted earth moving distance (WEMD) between the group distribution and the global distribution. Then we propose FedCGD algorithm to minimize the sum of sampling variance and WEMD on classification tasks by device scheduling and bandwidth allocation, within polynomial time. Simulation shows that the proposed strategy increases classification accuracy on the CIFAR-10 dataset by up to 4.2% while scheduling 41.8% fewer devices, and flexibly switches between reducing WEMD and reducing sampling variance.
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 27 (5) , 5713-5726
Energy-Efficient Collaborative Perception: A Block-Skipping DNN Approach With Dynamic Frequency Scaling and Environment Awareness
Energy efficiency is crucial for Connected Autonomous Vehicles (CAVs), where real-time perception via Deep Neural Networks (DNNs) demands significant computing resources. Although techniques such as Dynamic Frequency Scaling (DFS) and block-skipping reduce energy usage, they may degrade accuracy or increase inference speed. Integrating these methods with Collaborative Perception (CP), which leverages data sharing among vehicles to improve perception performances, offers a potential trade-off between accuracy and energy usage. This paper introduces EC-PUBSE (Efficient Collaborative Perception Using Block-Skipping, DFS, and Environment awareness), an energy-efficient CP framework for CAVs. EC-PUBSE dynamically allocates computing resources using block-skipping and DFS, coupled with environment-aware response time budget to sustain accuracy and reduce energy. We formulate the problem as a mixed-integer nonlinear programming (MINLP) problem, addressing the trade-off between computation energy and accuracy, under communication and response time budget constraints. We leverage model diversity and redundancy to improve system performance compared to standalone and CP. The proposed framework adapts to traffic conditions to offer an adaptable solution for sustainable autonomous driving. Experiments show that EC-PUBSE reduces energy consumption by up to 22%; it improves energy efficiency by up to $2.7\times $ and $2.3\times $ vs. standalone and collaborative only, respectively, maintaining perception performance in dynamic conditions.
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 27 (5) , 5713-5726
Energy-Efficient Collaborative Perception: A Block-Skipping DNN Approach With Dynamic Frequency Scaling and Environment Awareness
Energy efficiency is crucial for Connected Autonomous Vehicles (CAVs), where real-time perception via Deep Neural Networks (DNNs) demands significant computing resources. Although techniques such as Dynamic Frequency Scaling (DFS) and block-skipping reduce energy usage, they may degrade accuracy or increase inference speed. Integrating these methods with Collaborative Perception (CP), which leverages data sharing among vehicles to improve perception performances, offers a potential trade-off between accuracy and energy usage. This paper introduces EC-PUBSE (Efficient Collaborative Perception Using Block-Skipping, DFS, and Environment awareness), an energy-efficient CP framework for CAVs. EC-PUBSE dynamically allocates computing resources using block-skipping and DFS, coupled with environment-aware response time budget to sustain accuracy and reduce energy. We formulate the problem as a mixed-integer nonlinear programming (MINLP) problem, addressing the trade-off between computation energy and accuracy, under communication and response time budget constraints. We leverage model diversity and redundancy to improve system performance compared to standalone and CP. The proposed framework adapts to traffic conditions to offer an adaptable solution for sustainable autonomous driving. Experiments show that EC-PUBSE reduces energy consumption by up to 22%; it improves energy efficiency by up to $2.7\times $ and $2.3\times $ vs. standalone and collaborative only, respectively, maintaining perception performance in dynamic conditions.
- Open Access
Timeliness-Oriented Scheduling and Resource Allocation in Multi-Region Collaborative Perception
Collaborative perception (CP) is a critical technology in applications like autonomous driving and smart cities. It involves the sharing and fusion of information among sensors to overcome the limitations of individual perception, such as blind spots and range limitations. However, CP faces two primary challenges. First, due to the dynamic nature of the environment, the timeliness of the transmitted information is critical to perception performance. Second, with limited computational power at the sensors and constrained wireless bandwidth, the communication volume must be carefully designed to ensure feature representations are both effective and sufficient. This work studies the dynamic scheduling problem in a multi-region CP scenario, and presents a Timeliness-Aware Multi-region Prioritized (TAMP) scheduling algorithm to trade-off perception accuracy and communication resource usage. Timeliness reflects the utility of information that decays as time elapses, which is manifested by the perception performance in CP tasks. We propose an empirical penalty function that maps the joint impact of Age of Information (AoI) and communication volume to perception performance. Aiming to minimize this timeliness-oriented penalty in the long-term, and recognizing that scheduling decisions have a cumulative effect on subsequent system states, we propose the TAMP scheduling algorithm. TAMP is a Lyapunov-based optimization policy that decomposes the long-term average objective into a per-slot prioritization problem, balancing the scheduling worth against resource cost. We validate our algorithm in both intersection and corridor scenarios with the real-world Roadside Cooperative perception (RCooper) dataset. Extensive simulations demonstrate that TAMP outperforms the best-performing baseline, achieving an Average Precision (AP) improvement of up to 27% across various configurations.
Open AccessTimeliness-Oriented Scheduling and Resource Allocation in Multi-Region Collaborative Perception
Collaborative perception (CP) is a critical technology in applications like autonomous driving and smart cities. It involves the sharing and fusion of information among sensors to overcome the limitations of individual perception, such as blind spots and range limitations. However, CP faces two primary challenges. First, due to the dynamic nature of the environment, the timeliness of the transmitted information is critical to perception performance. Second, with limited computational power at the sensors and constrained wireless bandwidth, the communication volume must be carefully designed to ensure feature representations are both effective and sufficient. This work studies the dynamic scheduling problem in a multi-region CP scenario, and presents a Timeliness-Aware Multi-region Prioritized (TAMP) scheduling algorithm to trade-off perception accuracy and communication resource usage. Timeliness reflects the utility of information that decays as time elapses, which is manifested by the perception performance in CP tasks. We propose an empirical penalty function that maps the joint impact of Age of Information (AoI) and communication volume to perception performance. Aiming to minimize this timeliness-oriented penalty in the long-term, and recognizing that scheduling decisions have a cumulative effect on subsequent system states, we propose the TAMP scheduling algorithm. TAMP is a Lyapunov-based optimization policy that decomposes the long-term average objective into a per-slot prioritization problem, balancing the scheduling worth against resource cost. We validate our algorithm in both intersection and corridor scenarios with the real-world Roadside Cooperative perception (RCooper) dataset. Extensive simulations demonstrate that TAMP outperforms the best-performing baseline, achieving an Average Precision (AP) improvement of up to 27% across various configurations.
- 31 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications
- Vol. 32 (6) , 16-17
Guest Editorial: Special Issue on 6G Local Area Networks
- 31 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Wireless Communications
- Vol. 32 (6) , 16-17
Guest Editorial: Special Issue on 6G Local Area Networks
- Open Access
UniMM-V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Driving
Autonomous driving holds transformative potential but remains fundamentally constrained by the limited perception and isolated decision-making with standalone intelligence. While recent multi-agent approaches introduce cooperation, they often focus merely on perception-level tasks, overlooking the alignment with downstream planning and control, or fall short in leveraging the full capacity of the recent emerging end-to-end autonomous driving. In this paper, we present UniMM-V2X, a novel end-to-end multi-agent framework that enables hierarchical cooperation across perception, prediction, and planning. At the core of our framework is a multi-level fusion strategy that unifies perception and prediction cooperation, allowing agents to share queries and reason cooperatively for consistent and safe decision-making. To adapt to diverse downstream tasks and further enhance the quality of multi-level fusion, we incorporate a Mixture-of-Experts (MoE) architecture to dynamically enhance the BEV representations. We further extend MoE into the decoder to better capture diverse motion patterns. Extensive experiments on the DAIR-V2X dataset demonstrate our approach achieves state-of-the-art (SOTA) performance with a 39.7% improvement in perception accuracy, a 7.2% reduction in prediction error, and a 33.2% improvement in planning performance compared with UniV2X, showcasing the strength of our MoE-enhanced multi-level cooperative paradigm.
Open AccessUniMM-V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Driving
Autonomous driving holds transformative potential but remains fundamentally constrained by the limited perception and isolated decision-making with standalone intelligence. While recent multi-agent approaches introduce cooperation, they often focus merely on perception-level tasks, overlooking the alignment with downstream planning and control, or fall short in leveraging the full capacity of the recent emerging end-to-end autonomous driving. In this paper, we present UniMM-V2X, a novel end-to-end multi-agent framework that enables hierarchical cooperation across perception, prediction, and planning. At the core of our framework is a multi-level fusion strategy that unifies perception and prediction cooperation, allowing agents to share queries and reason cooperatively for consistent and safe decision-making. To adapt to diverse downstream tasks and further enhance the quality of multi-level fusion, we incorporate a Mixture-of-Experts (MoE) architecture to dynamically enhance the BEV representations. We further extend MoE into the decoder to better capture diverse motion patterns. Extensive experiments on the DAIR-V2X dataset demonstrate our approach achieves state-of-the-art (SOTA) performance with a 39.7% improvement in perception accuracy, a 7.2% reduction in prediction error, and a 33.2% improvement in planning performance compared with UniV2X, showcasing the strength of our MoE-enhanced multi-level cooperative paradigm.