Kun Qian
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
Kun Qian has a total of 76 co-authors. Here are the top co-authors by number of shared publications:
Fang Fang
14 shared publicationsXudong Ma
12 shared publications- Shan Luo
Department of Engineering, King's College London, London, U.K.
11 shared publications Bo Zhou
9 shared publications- Xingshuo Jing
School of Automation, Southeast University, Nanjing, China
9 shared publications - Bo Zhou
Southeast University
7 shared publications
Latest Publications
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 12  (1) , 33-40
DT-TacVLA: A Dual-Timescale Tactile-Enhanced Vision-Language-Action Model for Contact-Rich Manipulation
- 1 January 2027
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 12  (1) , 33-40
DT-TacVLA: A Dual-Timescale Tactile-Enhanced Vision-Language-Action Model for Contact-Rich Manipulation
- 5 November 2026
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 209
Geometry-aware depth completion of transparent objects in RGB-D optical sensing systems for robotic manipulation
- 5 November 2026
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 209
Geometry-aware depth completion of transparent objects in RGB-D optical sensing systems for robotic manipulation
- 10 September 2026
- Published by Emerald Publishing in Industrial Robot: the international journal of robotics research and application
- p. 1-14
Jamming-free reliable peg-in-hole assembly with unknown in-hand pose using visuotactile sensing
Purpose: This paper aims to propose a robust, jamming-free peg-in-hole assembly control method for complex components, such as threaded screws, in unstructured environments using visuotactile sensing. Design/methodology/approach: The approach introduces high-precision pose pre-alignment by performing joint estimation of the peg’s angle and offset from visuotactile depth images using a residual network. A dual-mask spatial averaging algorithm is formulated, providing a computationally efficient and low-latency solution for visuotactile force estimation. Furthermore, the search strategy integrates an active jamming recovery mechanism into an admittance-based spiral search, which leverages bidirectional transverse movements to release geometric constraints. Experiments were conducted using a 6-DOF robot arm equipped with visuotactile sensors to perform peg-in-hole tasks. Findings: The proposed framework achieved a 90% assembly success rate across 60 trials, with an average hole-searching time of 11.85 s. Time-domain response experiments demonstrated that our dual-mask force estimation algorithm effectively balances real-time responsiveness and noise suppression. In the step-response experiment, the proposed dual-mask force estimation algorithm achieved a low phase lag of 53.40 ms while maintaining high signal stability (Coefficient of Variation = 0.59%). Originality/value: This work contributes a complete jamming-free visuotactile assembly framework. By coupling a computationally efficient force estimation technique with a bidirectional escape mechanism, it provides a highly practical solution that enhances robotic assembly reliability.
- 10 September 2026
- Published by Emerald Publishing in Industrial Robot: the international journal of robotics research and application
- p. 1-14
Jamming-free reliable peg-in-hole assembly with unknown in-hand pose using visuotactile sensing
Purpose: This paper aims to propose a robust, jamming-free peg-in-hole assembly control method for complex components, such as threaded screws, in unstructured environments using visuotactile sensing. Design/methodology/approach: The approach introduces high-precision pose pre-alignment by performing joint estimation of the peg’s angle and offset from visuotactile depth images using a residual network. A dual-mask spatial averaging algorithm is formulated, providing a computationally efficient and low-latency solution for visuotactile force estimation. Furthermore, the search strategy integrates an active jamming recovery mechanism into an admittance-based spiral search, which leverages bidirectional transverse movements to release geometric constraints. Experiments were conducted using a 6-DOF robot arm equipped with visuotactile sensors to perform peg-in-hole tasks. Findings: The proposed framework achieved a 90% assembly success rate across 60 trials, with an average hole-searching time of 11.85 s. Time-domain response experiments demonstrated that our dual-mask force estimation algorithm effectively balances real-time responsiveness and noise suppression. In the step-response experiment, the proposed dual-mask force estimation algorithm achieved a low phase lag of 53.40 ms while maintaining high signal stability (Coefficient of Variation = 0.59%). Originality/value: This work contributes a complete jamming-free visuotactile assembly framework. By coupling a computationally efficient force estimation technique with a bidirectional escape mechanism, it provides a highly practical solution that enhances robotic assembly reliability.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73  (7) , 10229-10241
Explorative Grasping Referred Objects With Weak Stiffness via Vision-Language-Tactile Fusion
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73  (7) , 10229-10241
Explorative Grasping Referred Objects With Weak Stiffness via Vision-Language-Tactile Fusion
- 16 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. PP  (10834435) , 1-12
Learning Physics-Aware Sensorimotor Model With Visual–Tactile Sensing for DLO Manipulation
Robotic manipulation of deformable linear objects (DLOs)—including electrical cables, wires, and hoses—is prevalent in industrial applications. Achieving precise shape control in these tasks hinges on learning a reliable sensorimotor model that maps robot motion to resulting deformations. However, developing such models remains challenging, especially for DLOs with unknown or variable physical properties. To address this issue, we propose a physics-aware deformable linear object shaping (PA-DLOS) framework, which facilitates accurate shape control across a range of unseen DLOs. Specifically, the proposed neural network-based global DLO deformation model, trained in simulation, is conditioned on physics embeddings estimated in real time through a multimodal model that integrates visual, tactile, and proprioceptive inputs. On this basis, precise shape servoing of unseen DLOs can be achieved via a dual-stage online optimization strategy, which modulates the simulation-trained deformation model according to online visual and tactile feedback. Results from extensive simulation and real-world experiments demonstrate that PA-DLOS outperforms representative methods in both DLO deformation modeling and shape control. Notably, despite being trained solely in simulation, our framework achieves a 100% success rate, the highest tracking accuracy, and the fastest convergence in real-world shape servoing tasks.
- 16 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. PP  (10834435) , 1-12
Learning Physics-Aware Sensorimotor Model With Visual–Tactile Sensing for DLO Manipulation
Robotic manipulation of deformable linear objects (DLOs)—including electrical cables, wires, and hoses—is prevalent in industrial applications. Achieving precise shape control in these tasks hinges on learning a reliable sensorimotor model that maps robot motion to resulting deformations. However, developing such models remains challenging, especially for DLOs with unknown or variable physical properties. To address this issue, we propose a physics-aware deformable linear object shaping (PA-DLOS) framework, which facilitates accurate shape control across a range of unseen DLOs. Specifically, the proposed neural network-based global DLO deformation model, trained in simulation, is conditioned on physics embeddings estimated in real time through a multimodal model that integrates visual, tactile, and proprioceptive inputs. On this basis, precise shape servoing of unseen DLOs can be achieved via a dual-stage online optimization strategy, which modulates the simulation-trained deformation model according to online visual and tactile feedback. Results from extensive simulation and real-world experiments demonstrate that PA-DLOS outperforms representative methods in both DLO deformation modeling and shape control. Notably, despite being trained solely in simulation, our framework achieves a 100% success rate, the highest tracking accuracy, and the fastest convergence in real-world shape servoing tasks.
- Open Access
Real-Time LiDAR 3D Semantic Segmentation via Multi-View and Cross-Modal Compact Featuring Two-Branch Knowledge Distillation
Simultaneous online mapping and semantic segmentation using handheld scanners supports various environmental inspection and measurement tasks. For such scanners, combing visual and LiDAR data is beneficial for improving the segmentation performance. But the direct fusion of multi-modal and multi-view features faces challenges in terms of both real-time performance and robustness. To address these challenges, this paper proposes a multi-view and cross-modal knowledge distillation method for supporting runtime LiDAR-only semantic segmentation. The proposed method hierarchically compacts multi-view and cross-model priors and distills them into two branches to improve segmentation accuracy. In addition, we design an improved data augmentation technique based on PolarMix for rendering more realistic point cloud scenes. The experimental results on the SemanticKITTI and nuScenes datasets demonstrate that the mIoU of our approach outperforms the state-of-the-art knowledge-distillation-based methods. In addition, mapping experiments using a handheld scanner demonstrate the proposed method’s superior real-time performance and accuracy.
Open AccessReal-Time LiDAR 3D Semantic Segmentation via Multi-View and Cross-Modal Compact Featuring Two-Branch Knowledge Distillation
Simultaneous online mapping and semantic segmentation using handheld scanners supports various environmental inspection and measurement tasks. For such scanners, combing visual and LiDAR data is beneficial for improving the segmentation performance. But the direct fusion of multi-modal and multi-view features faces challenges in terms of both real-time performance and robustness. To address these challenges, this paper proposes a multi-view and cross-modal knowledge distillation method for supporting runtime LiDAR-only semantic segmentation. The proposed method hierarchically compacts multi-view and cross-model priors and distills them into two branches to improve segmentation accuracy. In addition, we design an improved data augmentation technique based on PolarMix for rendering more realistic point cloud scenes. The experimental results on the SemanticKITTI and nuScenes datasets demonstrate that the mIoU of our approach outperforms the state-of-the-art knowledge-distillation-based methods. In addition, mapping experiments using a handheld scanner demonstrate the proposed method’s superior real-time performance and accuracy.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (3) , 2434-2441
PoCoDP3: Pose- and Contact-Aware Visual-Tactile Policy for Contact-Rich 3D Manipulation
Imitation learning in contact-rich tasks requires both global spatial awareness and fine-grained in-hand interaction understanding. However, vision-only policies based on images or point clouds are often susceptible to occlusion and struggle to capture critical contact details, particularly in visually ambiguous regions or during subtle tactile interactions. In this work, we present PoCoDP3, a pose- and contact-aware visual-tactile policy that integrates 3D point clouds and tactile inputs to generate actions in contact-rich tasks. PoCoDP3 introduces a dual-branch tactile encoder that jointly models contact dynamics and estimates in-hand object pose, enabling structured tactile representations for precise contact-rich manipulation. A contact-driven cross-modal fusion mechanism adaptively prioritizes sensory modalities based on real-time interaction cues, enabling efficient visual-tactile integration. Moreover, a reference-guided diffusion policy leverages reference action offsets to reduce sampling steps, significantly accelerating inference while maintaining action quality. Experiments across simulation and real-world tasks demonstrate that PoCoDP3 consistently outperforms representative 2D and 3D policies in terms of both accuracy and inference efficiency.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11  (3) , 2434-2441
PoCoDP3: Pose- and Contact-Aware Visual-Tactile Policy for Contact-Rich 3D Manipulation
Imitation learning in contact-rich tasks requires both global spatial awareness and fine-grained in-hand interaction understanding. However, vision-only policies based on images or point clouds are often susceptible to occlusion and struggle to capture critical contact details, particularly in visually ambiguous regions or during subtle tactile interactions. In this work, we present PoCoDP3, a pose- and contact-aware visual-tactile policy that integrates 3D point clouds and tactile inputs to generate actions in contact-rich tasks. PoCoDP3 introduces a dual-branch tactile encoder that jointly models contact dynamics and estimates in-hand object pose, enabling structured tactile representations for precise contact-rich manipulation. A contact-driven cross-modal fusion mechanism adaptively prioritizes sensory modalities based on real-time interaction cues, enabling efficient visual-tactile integration. Moreover, a reference-guided diffusion policy leverages reference action offsets to reduce sampling steps, significantly accelerating inference while maintaining action quality. Experiments across simulation and real-world tasks demonstrate that PoCoDP3 consistently outperforms representative 2D and 3D policies in terms of both accuracy and inference efficiency.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22  (4) , 3585-3596
ViTaDex: Vision–Tactile Fusion for 6-D Object-in-Hand Pose Estimation in Dexterous Anthropomorphic Manipulation
Object-in-hand 6-D pose estimation is critical for dexterous manipulation and assembly tasks. Existing object-in-hand pose estimation methods neglect the varying significance of tactile information from distinct hand regions during feature extraction and employ simplistic concatenation for vision–tactile fusion, which fails to capture the complementary relationships between these modalities. Moreover, prevailing object-in-hand datasets predominantly rely on predefined motion trajectories and single-point fingertip contact information, which inadequately represent real-world scenarios. To address these limitations, we propose the vision–tactile adaptive fusion network. Specifically, we design a novel tactile-gated adaptive graph convolution to dynamically model multiregional tactile features across the hand. We also introduce a vision–tactile cross-attention gated fusion module that facilitates efficient integration of multimodal sensory information. In addition, we construct the Vision–Tactile Object-in-Hand Dexterous Manipulation dataset based on human demonstration for 6-D pose estimation. The dataset employs a teleoperation system that combines an exoskeleton hand with a pose tracking system to replicate human dexterous behaviors and encompasses high-density tactile data spanning the entire hand. Extensive experiments demonstrate the robustness of our framework against occlusions.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22  (4) , 3585-3596
ViTaDex: Vision–Tactile Fusion for 6-D Object-in-Hand Pose Estimation in Dexterous Anthropomorphic Manipulation
Object-in-hand 6-D pose estimation is critical for dexterous manipulation and assembly tasks. Existing object-in-hand pose estimation methods neglect the varying significance of tactile information from distinct hand regions during feature extraction and employ simplistic concatenation for vision–tactile fusion, which fails to capture the complementary relationships between these modalities. Moreover, prevailing object-in-hand datasets predominantly rely on predefined motion trajectories and single-point fingertip contact information, which inadequately represent real-world scenarios. To address these limitations, we propose the vision–tactile adaptive fusion network. Specifically, we design a novel tactile-gated adaptive graph convolution to dynamically model multiregional tactile features across the hand. We also introduce a vision–tactile cross-attention gated fusion module that facilitates efficient integration of multimodal sensory information. In addition, we construct the Vision–Tactile Object-in-Hand Dexterous Manipulation dataset based on human demonstration for 6-D pose estimation. The dataset employs a teleoperation system that combines an exoskeleton hand with a pose tracking system to replicate human dexterous behaviors and encompasses high-density tactile data spanning the entire hand. Extensive experiments demonstrate the robustness of our framework against occlusions.
- 21 October 2025
- Published by Elsevier in Journal of Manufacturing Processes
- Vol. 155, 428-442
Polishing via ODEs: Adaptive admittance control for robot polishing based on Neural ODEs
- 21 October 2025
- Published by Elsevier in Journal of Manufacturing Processes
- Vol. 155, 428-442
Polishing via ODEs: Adaptive admittance control for robot polishing based on Neural ODEs