👋 About Me

Hi 👋! I am Kaiyue Chen, an undergraduate student in Mechatronics and Robotic Systems at Xi’an Jiaotong-Liverpool University (XJTLU) in Suzhou, China. My research focuses on tactile-guided Vision-Language-Action (VLA), embodied manipulation.

I am currently a research intern at Sudo Robotics, advised by Prof. Rui Chen, where I develop tactile-guided VLA systems for contact-rich bimanual manipulation. My work includes vision-based tactile sensing, a dual-arm SmolVLA pipeline, tactile video encoding, and Flow Reversal Steering for action correction.

Previously, I co-organized the ManiSkill-ViTac Challenge 2026 at the CVPR 2026 Embodied AI Workshop and developed its tactile-VLA baseline. I am also a co-first author of 3WD-VLA, a three-way decision framework for safer robotic grasping under ambiguity.

🔥 News

• 2026.03:  🤖 Joined Sudo Robotics as a research intern working on tactile-guided VLA for contact-rich manipulation.
• 2026.06:  🥳 ManiSkill-ViTac Challenge 2026 was successfully completed.
• 2026:  📝 Submitted two papers on Three-Way Decision VLA and real-time robotic grasp detection.
• 2025.08:  🏆 National Finals, The 27th China Robot and Artificial Intelligence Competition — Baidu Service Robot track, coached by Prof. Eng Gee Lim (IEEE Fellow).
• 2024.09:  📚 Started BEng Mechatronics and Robotic Systems at XJTLU (expected 2028).

📝 Publications

^* Equal contribution.

• Yan Zhu^*, Kaiyue Chen^*, Lingyu Li, Xiaohui Zhu, and Xinheng Wang. “3WD-VLA: Hierarchical Three-Way Decision for Collaborative Dual-View Robotic Grasping.” Submitted to the 22nd EAI International Conference on Collaborative Computing: Networking, Applications and Worksharing (EAI CollaborateCom 2026).

• Xiaoyu Xia, Chengcheng Hu, Leyi Liu, Junyan He, Kaiyue Chen, Yuhao Jin, Yong Yue, and Xiaohui Zhu. “Real-Time 3-DoF Robotic Grasp Detection via Keypoint Regression.” Submitted to the 31st International Conference on Automation and Computing (ICAC 2026).

💬 Research Experience

• 2026.06 - Present, Research Intern, Tactile-Guided VLA for Contact-Rich Robotic Manipulation, Sudo Robotics (Advisor: Prof. Rui Chen).
  • Designed and deployed vision-based tactile sensors on bimanual grippers to mitigate partial observability caused by occlusion and limited visual feedback.
  • Built a dual-arm tactile-VLA pipeline based on SmolVLA (0.7B) and am developing a tactile video encoder for contact geometry, shear deformation, and slip-related dynamics.
  • Explored Flow Reversal Steering to map tactile-guided coarse action corrections into the latent noise space of pretrained VLA policies.
• 2026.03 - 2026.06, Student Organizer, ManiSkill-ViTac Challenge 2026, CVPR 2026 Embodied AI Workshop.
  • Built an end-to-end VLA pipeline around UMI-based data collection, covering hardware maintenance, data processing and filtering, model training, and real-robot deployment.
  • Integrated AnyTouch tactile representations into a π0.5-based VLA using modality-weighted training to establish the competition baseline.
  • Designed two long-horizon, contact-rich tasks and co-organized the challenge and accompanying technical report.
• 2026.01 - 2026.05, Co-first Author, Three-Way Decision VLA, Xi’an Jiaotong-Liverpool University (Advisor: Prof. Xinheng Wang).
  • Developed a collaborative robotic grasping framework that defers action when direct execution is unsafe or unreliable, including ambiguous commands, target absence, human-hand intrusion, occlusion, and stacked objects.
  • Built a teleoperation data-collection pipeline for the SO-100 robot arm and trained ACT, π0.5, and SmolVLA models.
  • Improved performance by 27.5 percentage points over a matched SmolVLA baseline without 3WD and demonstrated clearer behavior under ambiguity, occlusion, and object stacking.

💼 Internships

Sudo Robotics

2026.06 - Present · Research Intern

Tactile Sensing VLA Bimanual Manipulation

🧩 Projects

ManiSkill-ViTac Challenge 2026: Real-world Language-guided Bimanual Vision-Tactile Manipulation

CVPR 2026 Embodied AI Workshop · Student Organizer

CVPR 2026 VLA Tactile

Project Page

💻 Competitions

• 2025.08, National Finals Team Member, Baidu Service Robot Competition, The 27th China Robot and Artificial Intelligence Competition, Suzhou; coached by Prof. Eng Gee Lim (IEEE Fellow).
  • Developed a ROS-based autonomous restaurant-service robot with order-driven task coordination, integrating touchscreen interaction, multi-point navigation, food pickup, delivery, and checkout.

📖 Education

• 2024.09 - 2028.07 (expected), BEng Mechatronics and Robotic Systems, Xi’an Jiaotong-Liverpool University, Suzhou, China. GPA (Years 1-2): 3.83/4.0. Core coursework: Linear Algebra (4.0), Calculus (4.0), and C Programming (4.0).

🛠 Technical Skills

• Robotics: PyTorch, deep reinforcement learning, ROS1/2, Gazebo, Isaac Sim, ManiSkill.
• Programming: Python, C, C++.
• Design and tools: SolidWorks, Autodesk Fusion 360.