Wenchao Ding
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
Research Interests
Wenchao Ding has a total of 155 co-authors. Here are the top co-authors by number of shared publications:
- Jieru Zhao
Shanghai Jiao Tong University, Shanghai, China
9 shared publications - Zhongxue Gan
Fudan University
8 shared publications Zhongxue Gan
Fudan University
7 shared publications- Minyi Guo
Shanghai Jiao Tong University, Shanghai, China
5 shared publications Kaizhao Zhang
4 shared publications- Quan Chen
Shanghai Jiao Tong University
4 shared publications
Latest Publications
- 9 September 2026
- Published by Springer Nature
- p. 37-53
AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation
- 9 September 2026
- Published by Springer Nature
- p. 37-53
AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation
- 1 August 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (8) , 10042-10049
Hammer: H ierarchical A ttention Generation M eets M ixture-of- E xpe r ts Evaluation for End-to-End Autonomous Driving
- 1 August 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (8) , 10042-10049
Hammer: H ierarchical A ttention Generation M eets M ixture-of- E xpe r ts Evaluation for End-to-End Autonomous Driving
- Open Access
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
LiveVLM: Efficient Online Video Understanding via Streaming-Oriented KV Cache and Retrieval
Recent developments in Video Large Language Models (Video LLMs) have enabled models to process hour-long videos and exhibit exceptional performance. Nonetheless, the Key-Value (KV) cache expands linearly over time, leading to substantial memory overhead and response delay-critical challenges in various real-world online applications, such as Deepseek services, autonomous driving and robotics. To mitigate these issues, we propose LiveVLM, a training-free and query-agnostic framework specifically designed for online video understanding and real-time interaction. LiveVLM employs a Vision Sink Bucketing (VSB) mechanism to process video streams in real time, retain long-term video details and eliminate redundant KVs. This mechanism utilizes vision-to-vision attention scores as the metric and seeks to maximize the coverage of contextual information during compression. Noting that KV cache compressed in a query-agnostic manner inevitably retains irrelevant information for specific queries, LiveVLM incorporates a Position-agnostic KV Retrieval (PaR) mechanism to reduce interference from redundant context. The keypoint of PaR lies in decoupling positional embeddings to enhance the similarity between key tensors, thereby supporting efficient retrieval at the granularity of pages. Extensive experiments demonstrate that LiveVLM enables the foundation LLaVA-OneVision model to achieve state-of-the-art accuracy among both training-free query-agnostic methods and training-based online models. Our code is available at https://github.com/sjtu-zhao-lab/LiveVLM.
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessLiveVLM: Efficient Online Video Understanding via Streaming-Oriented KV Cache and Retrieval
Recent developments in Video Large Language Models (Video LLMs) have enabled models to process hour-long videos and exhibit exceptional performance. Nonetheless, the Key-Value (KV) cache expands linearly over time, leading to substantial memory overhead and response delay-critical challenges in various real-world online applications, such as Deepseek services, autonomous driving and robotics. To mitigate these issues, we propose LiveVLM, a training-free and query-agnostic framework specifically designed for online video understanding and real-time interaction. LiveVLM employs a Vision Sink Bucketing (VSB) mechanism to process video streams in real time, retain long-term video details and eliminate redundant KVs. This mechanism utilizes vision-to-vision attention scores as the metric and seeks to maximize the coverage of contextual information during compression. Noting that KV cache compressed in a query-agnostic manner inevitably retains irrelevant information for specific queries, LiveVLM incorporates a Position-agnostic KV Retrieval (PaR) mechanism to reduce interference from redundant context. The keypoint of PaR lies in decoupling positional embeddings to enhance the similarity between key tensors, thereby supporting efficient retrieval at the granularity of pages. Extensive experiments demonstrate that LiveVLM enables the foundation LLaVA-OneVision model to achieve state-of-the-art accuracy among both training-free query-agnostic methods and training-based online models. Our code is available at https://github.com/sjtu-zhao-lab/LiveVLM.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (7) , 8439-8446
Resolving State Ambiguity in Robot Manipulation via Adaptive Working Memory Recoding
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (7) , 8439-8446
Resolving State Ambiguity in Robot Manipulation via Adaptive Working Memory Recoding
- Open Access
- 26 June 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 23  (2) , 1-26
NPUMeter: Automatic Operator Optimization for Ascend NPU with Accurate Analytical Performance Models
With the rapid development of AI and deep learning, computational demands are increasing significantly. While GPUs excel in parallel computing, they fall short in terms of energy efficiency, specialization, and processing latency. In contrast, Neural Processing Units (NPUs), such as the Ascend NPUs, designed specifically for deep learning tasks, demonstrate superior performance. However, the architecture specialization makes operator development more challenging, leading to a reliance on manual tuning and optimization, which incurs significant time cost and developing effort. To address this issue, we propose NPUMeter, an automatic operator optimization framework for Ascend NPUs built upon accurate and comprehensive analytical performance models. NPUMeter comprises two components: (1) an analytical performance model that accurately estimates operator latency on NPU given different configurations of optimization parameters; (2) an efficient design space exploration (DSE) algorithm that automatically searches for the optimal parameter configuration in a large design space within minutes. Experimental results demonstrate that NPUMeter achieves high estimation accuracy, with an average error below 5%. It effectively generates near-optimal configurations for various operators, achieving up to a 1.46× performance speedup compared to the configuration generated by the Ascend C compiler while reducing the DSE time from hours to minutes.
- 26 June 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 23  (2) , 1-26
Open AccessNPUMeter: Automatic Operator Optimization for Ascend NPU with Accurate Analytical Performance Models
With the rapid development of AI and deep learning, computational demands are increasing significantly. While GPUs excel in parallel computing, they fall short in terms of energy efficiency, specialization, and processing latency. In contrast, Neural Processing Units (NPUs), such as the Ascend NPUs, designed specifically for deep learning tasks, demonstrate superior performance. However, the architecture specialization makes operator development more challenging, leading to a reliance on manual tuning and optimization, which incurs significant time cost and developing effort. To address this issue, we propose NPUMeter, an automatic operator optimization framework for Ascend NPUs built upon accurate and comprehensive analytical performance models. NPUMeter comprises two components: (1) an analytical performance model that accurately estimates operator latency on NPU given different configurations of optimization parameters; (2) an efficient design space exploration (DSE) algorithm that automatically searches for the optimal parameter configuration in a large design space within minutes. Experimental results demonstrate that NPUMeter achieves high estimation accuracy, with an average error below 5%. It effectively generates near-optimal configurations for various operators, achieving up to a 1.46× performance speedup compared to the configuration generated by the Ascend C compiler while reducing the DSE time from hours to minutes.
- 19 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (4) , 4457-4464
Learning a Unified Risk Map for Autonomous Driving in Partially Observable Environments
Occlusion-aware prediction remains a critical challenge in autonomous driving due to the inherent uncertainty of unobserved regions. Existing approaches either overestimate risk based on reachable states or struggle to predict accurate trajectories under high occlusion uncertainty. To address these limitations, we propose a unified risk map modeling and learning framework for partially observable environments. Our method integrates traffic flow risk and collision risk through spatiotemporal modeling, enabling fine-grained assessment of occlusion-induced hazards. To address the scarcity of scenarios involving occluded interactions, we introduce a diffusion-based scenario generation framework that produces realistic yet adversarial scenarios. We integrate the modeling and learning of a unified risk map into a framework that supports risk-aware planning under partial observability. Experiments on the Waymo Open Motion Dataset show that our method significantly outperforms the state-of-the-art occlusion-aware baseline, improving minimum time-to-collision by 0.78 times and average time-to-collision by 1.67 times. The proposed framework offers a comprehensive and practical solution for risk-aware planning in partially observable environments.
- 19 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (4) , 4457-4464
Learning a Unified Risk Map for Autonomous Driving in Partially Observable Environments
Occlusion-aware prediction remains a critical challenge in autonomous driving due to the inherent uncertainty of unobserved regions. Existing approaches either overestimate risk based on reachable states or struggle to predict accurate trajectories under high occlusion uncertainty. To address these limitations, we propose a unified risk map modeling and learning framework for partially observable environments. Our method integrates traffic flow risk and collision risk through spatiotemporal modeling, enabling fine-grained assessment of occlusion-induced hazards. To address the scarcity of scenarios involving occluded interactions, we introduce a diffusion-based scenario generation framework that produces realistic yet adversarial scenarios. We integrate the modeling and learning of a unified risk map into a framework that supports risk-aware planning under partial observability. Experiments on the Waymo Open Motion Dataset show that our method significantly outperforms the state-of-the-art occlusion-aware baseline, improving minimum time-to-collision by 0.78 times and average time-to-collision by 1.67 times. The proposed framework offers a comprehensive and practical solution for risk-aware planning in partially observable environments.
- Open Access
Enhancing Indoor Occupancy Prediction via Sparse Query-Based Multi-Level Consistent Knowledge Distillation
Occupancy prediction provides critical geometric and semantic understanding for robotics but faces efficiency-accuracy trade-offs. Current dense methods suffer computational waste on empty voxels, while sparse query-based approaches lack robustness in diverse and complex indoor scenes. In this paper, we propose DiScene, a novel sparse query-based framework that leverages multi-level distillation to achieve efficient and robust occupancy prediction. In particular, our method incorporates two key innovations: (1) a Multi-level Consistent Knowledge Distillation strategy, which transfers hierarchical representations from large teacher models to lightweight students through coordinated alignment across four levels, including encoder-level feature alignment, query-level feature matching, prior-level spatial guidance, and anchor-level high-confidence knowledge transfer and (2) a Teacher-Guided Initialization policy, employing optimized parameter warm-up to accelerate model convergence. Validated on the Occ-Scannet benchmark, DiScene achieves 23.2 FPS without depth priors while outperforming our baseline method, OPUS, by 36.1% and even better than the depth-enhanced version, OPUS†. With depth integration, DiScene†attains new SOTA performance, surpassing EmbodiedOcc by 3.7% with 1.62$\times$ faster inference speed. Furthermore, experiments on the Occ3D-nuScenes benchmark and in-the-wild scenarios demonstrate the versatility of our approach in various environments. Code and models can be accessed at https://github.com/getterupper/DiScene.
Open AccessEnhancing Indoor Occupancy Prediction via Sparse Query-Based Multi-Level Consistent Knowledge Distillation
Occupancy prediction provides critical geometric and semantic understanding for robotics but faces efficiency-accuracy trade-offs. Current dense methods suffer computational waste on empty voxels, while sparse query-based approaches lack robustness in diverse and complex indoor scenes. In this paper, we propose DiScene, a novel sparse query-based framework that leverages multi-level distillation to achieve efficient and robust occupancy prediction. In particular, our method incorporates two key innovations: (1) a Multi-level Consistent Knowledge Distillation strategy, which transfers hierarchical representations from large teacher models to lightweight students through coordinated alignment across four levels, including encoder-level feature alignment, query-level feature matching, prior-level spatial guidance, and anchor-level high-confidence knowledge transfer and (2) a Teacher-Guided Initialization policy, employing optimized parameter warm-up to accelerate model convergence. Validated on the Occ-Scannet benchmark, DiScene achieves 23.2 FPS without depth priors while outperforming our baseline method, OPUS, by 36.1% and even better than the depth-enhanced version, OPUS†. With depth integration, DiScene†attains new SOTA performance, surpassing EmbodiedOcc by 3.7% with 1.62$\times$ faster inference speed. Furthermore, experiments on the Occ3D-nuScenes benchmark and in-the-wild scenarios demonstrate the versatility of our approach in various environments. Code and models can be accessed at https://github.com/getterupper/DiScene.
- Open Access
- 1 January 2026
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 147  (1) , 1-10
Planning by Simulation: A Query-Centric Search-Based Framework for Interactive Planning in Autonomous Driving
Ensuring operational safety for autonomous vehicles is a critical challenge in modern engineering, particularly due to the intricate interactions among diverse traffic participants. Traditional approaches often treat planning and prediction as unidirectional processes, failing to capture the dynamic, game-theoretic nature of real-world traffic. In the context of Digital Twins, there is an urgent need for high-fidelity virtual representations that can model the continuous, bidirectional evolution of the ego vehicle and surrounding agents to support robust decision-making under uncertainty. To address these limitations, a novel framework named Planning by Simulation with mutual influence prediction is proposed, which functions as a high-fidelity simulation-based predictive planner for autonomous driving decision-making. This framework explicitly models the iterative interplay between the ego vehicle’s planning and the predicted trajectories of surrounding agents within a virtual environment. By integrating a query-centric trajectory prediction mechanism with Monte Carlo Tree Search, the proposed approach orchestrates intelligent model exploration. It iteratively refines the ego vehicle’s actions by simulating future scenarios and adapting to the dynamic behaviors of other agents, thereby tightly coupling data-driven predictions with physics-based planning constraints. Comprehensive evaluations on the Argoverse 1 and Argoverse 2 dataset in Metadrive simulator demonstrate the efficacy of this simulation-based approach. The framework successfully captures complex interaction dynamics that static models overlook. The results indicate that the proposed method generates significantly safer, more rational, and human-like trajectories compared to existing baselines, validating the system’s high-fidelity predictive capabilities. The proposed framework illustrates the transformative potential of advanced virtual simulation technologies in autonomous mobility. By enabling the continuous integration of predictive data into the planning loop, this study provides a powerful foundation for interpretable and reliable decision-making in virtual engineering systems. It highlights how coupling generative simulation with interactive planning can resolve critical safety challenges in the lifecycle of intelligent autonomous systems.
- 1 January 2026
- Published by Tech Science Press in Computer Modeling in Engineering & Sciences
- Vol. 147  (1) , 1-10
Open AccessPlanning by Simulation: A Query-Centric Search-Based Framework for Interactive Planning in Autonomous Driving
Ensuring operational safety for autonomous vehicles is a critical challenge in modern engineering, particularly due to the intricate interactions among diverse traffic participants. Traditional approaches often treat planning and prediction as unidirectional processes, failing to capture the dynamic, game-theoretic nature of real-world traffic. In the context of Digital Twins, there is an urgent need for high-fidelity virtual representations that can model the continuous, bidirectional evolution of the ego vehicle and surrounding agents to support robust decision-making under uncertainty. To address these limitations, a novel framework named Planning by Simulation with mutual influence prediction is proposed, which functions as a high-fidelity simulation-based predictive planner for autonomous driving decision-making. This framework explicitly models the iterative interplay between the ego vehicle’s planning and the predicted trajectories of surrounding agents within a virtual environment. By integrating a query-centric trajectory prediction mechanism with Monte Carlo Tree Search, the proposed approach orchestrates intelligent model exploration. It iteratively refines the ego vehicle’s actions by simulating future scenarios and adapting to the dynamic behaviors of other agents, thereby tightly coupling data-driven predictions with physics-based planning constraints. Comprehensive evaluations on the Argoverse 1 and Argoverse 2 dataset in Metadrive simulator demonstrate the efficacy of this simulation-based approach. The framework successfully captures complex interaction dynamics that static models overlook. The results indicate that the proposed method generates significantly safer, more rational, and human-like trajectories compared to existing baselines, validating the system’s high-fidelity predictive capabilities. The proposed framework illustrates the transformative potential of advanced virtual simulation technologies in autonomous mobility. By enabling the continuous integration of predictive data into the planning loop, this study provides a powerful foundation for interpretable and reliable decision-making in virtual engineering systems. It highlights how coupling generative simulation with interactive planning can resolve critical safety challenges in the lifecycle of intelligent autonomous systems.