GBC: Generalized Behavior Cloning for Humanoid Robots
Yao, Y., Luo, C., Du, J., & Lu, J. (2024). "GBC: Generalized Behavior Cloning for Humanoid Robots." arXiv preprint arXiv:2508.09960.
CV
Personal Statement
My passion lies at the intersection of robotic embodiment and artificial intelligence, and my goal is to pioneer the next generation of autonomous robots capable of learning with human-like adaptability. As a Master’s student at SJTU, I have built a solid foundation through research and collaboration spanning reinforcement learning, robotic control, computer vision, and language models, and have applied these skills to real-world challenges from commercialized systems to humanoid control. Eager to deepen my expertise and pursue innovative solutions, I am seeking a PhD opportunity to contribute to research on creating robots that can learn from and interact with the world in a more human-like manner.
Education
- M.Sc. Control Science and Engineering, Shanghai Jiao Tong University (2023.9 - Present)
- School of Electronic Information and Electric Engineering (SEIEE)
- GPA: 3.89/4.0
- Machine Vision and Autonomous Systems Lab
- Advisor: Prof. Jun-Guo Lu
- B.Sc. Automation, Shanghai Jiao Tong University (2019.9 - 2023.6)
- School of Electronic Information and Electric Engineering (SEIEE)
- Zhiyuan Honors Program
- GPA: 3.65/4.0 (Major)
- Zhiyuan Honored Bachelor degree recipient
- Outstanding Graduate of Shanghai Jiao Tong University
- B.Sc. Economics (Minor), Shanghai Jiao Tong University (2019.9 - 2023.6)
- Antai College of Economics and Management
- Interdisciplinary program combining technical automation with economic analysis
Research & Work Experience
- Research Intern, Baosight Co., Ltd. (2024.9 - 2025.8)
- Developed and deployed humanoid robot reinforcement learning algorithms and designed imitation learning systems
- Successfully demonstrated humanoid robot capabilities of locomotion and manipulation at WAIC 2025, Shanghai
- Graduate Research Assistant, Machine Vision and Autonomous System Laboratory (2024.12 - Present)
- Conducting research on reinforcement learning and imitation learning for humanoid robots
- Supervised by Prof. Junguo Lu at Shanghai Jiao Tong University
- Focus: GBC framework development and cross-morphology learning
- Co-working and Product Testing, LIMX Dynamics (2024.6 - 2024.10)
- Testing LIMX P1 Series bipedal robots and shifting research focus from industrial CV to robotics
- Supervised by Prof. Junguo Lu; Collaboration with LIMX Dynamics, Shenzhen, China
- Applied techno-economic analysis to evaluate market positioning strategies
- Algorithm Developer, ZIDAI Industrial Collaboration (2023.11 - 2024.6)
- Designed Bird’s Eye View (BEV) generation system for ship-based surrounding view and warning systems
- Fine-tuned and deployed semantic segmentation and distance prediction models
- Product successfully commercialized with assessment of commercial viability and market penetration
- Mobile Application Developer, SAIC Motor AI Research Institute (2023.6 - 2023.11)
- Developed Android application for auto-calibration and surround view display in small vehicles
- Implemented Kotlin, JNI, OpenGL programming and user interface design
- Deployed in production vehicles for enhanced driver assistance
- Collaborative Research Assistant, Shanghai University of Sport (2022.11 - 2023.6)
- Designed and trained transformer-based models for human pose estimation and sequence analysis
- Developed integrated analysis system with GUI; system currently deployed and in active use
- Supporting multiple Olympic sports training programs
- National Undergraduate Innovative Test Program, Vision and Lidar Integrated Navigation (2020.11 - 2022.11)
- Participated in micro vehicle navigation research under supervision of Prof. Jingchuan Wang
- Gained foundational experience in sensor fusion and autonomous navigation systems
Technical Skills
- Programming Languages
- C/C++ (Advanced), Python (Advanced), MATLAB, LaTeX, Shell Scripting, Java
- Deep Learning & AI
- PyTorch, TensorFlow, JAX, Weights & Biases, Isaac Gym, Isaac Lab, OpenAI Gym
- Robotics & Simulation
- ROS/ROS2, Gazebo, MuJoCo, Isaac Sim, OpenCV, OpenGL
- Development Tools
- Git/GitHub, Docker, Qt, CMake, Linux/Ubuntu, VS Code
- Hardware & Embedded
- FPGA Development, Embedded Systems, Microcontrollers, Hardware-Software Integration
- Languages
- English (Fluent), Mandarin Chinese (Native), Japanese (Basic Reading)
Research Vision and Interests
Core Research Interests
- Embodied Artificial Intelligence: Foundation Models for Robotics, Language-Grounded Agents, Hierarchical Reinforcement & Imitation Learning
- Generative Models for Motion: Text-to-Motion Synthesis, Diffusion Models for Trajectory Planning, World Models
- Autonomous Systems: End-to-End Robot Learning, Sim-to-Real Transfer, Research Automation Workflows
Research Contributions & Technical Details
AnyBipe Framework: Developed a fully automated, end-to-end pipeline to streamline the entire robot learning workflow. This framework accelerates research by leveraging LLMs for automated reward design, training supervision, and sim-to-real validation.
GBC Framework: Engineered a universal imitation learning system enabling heterogeneous humanoids to acquire complex motor skills from motion data. Key innovations include a differentiable IK network for zero-shot motion retargeting and a novel DAgger-MMPPO algorithm. Open-sourced at https://github.com/sjtu-mvasl-robotics/GBC.
GBC Diffusion: Pioneered a text-driven, morphology-adaptive diffusion model that directly generates physically-plausible actions, serving as a high-level planner. By training directly on retargeted datasets, this model generates long-horizon motions (8s) in 50-100ms.
Awards & Achievements
- RoboCup 2021: First Prize - International robotics competition
- MCM/ICM 2021: Outstanding Winner (Meritorious Winner) - Mathematical Contest in Modeling
- First Prize Scholarship: Master’s program (2023-2026) - Academic excellence recognition
- Zhiyuan Honored Scholarship: Undergraduate program (2019-2023) - Top-tier academic performance
Publications
AnyBipe: An End-to-End Framework for Training and Deploying Bipedal Robots Guided by Large Language Models
Yao, Y., He, W., Gu, C., Du, J., Tan, F., Zhu, Z., & Lu, J. (2025). "AnyBipe: An End-to-End Framework for Training and Deploying Bipedal Robots Guided by Large Language Models." Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Oral Presentation.
Publications
Published & Accepted:
- Yao, Y., He, W., Gu, C., Du, J., Tan, F., Zhu, Z., & Lu, J. (2025). “AnyBipe: An End-to-End Framework for Training and Deploying Bipedal Robots Guided by Large Language Models.” Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). [Oral Presentation]
Published Preprints:
- Yao, Y., Luo, C., Du, J., & Lu, J. (2024). “GBC: Generalized Behavior Cloning for Humanoid Robots.” arXiv preprint arXiv:2508.09960.
In Progress:
- Continuing research on multimodal learning frameworks for autonomous humanoid systems (will be released when accepted)
Awards & Achievements
- Academic Scholarship (Tier 1), Shanghai Jiao Tong University (2024, 2025)
- Merit-based scholarship for exceptional academic performance during graduate studies
- Best Student Presentation Award, China Association of Automation (CAA), Shanghai (2023, 2024)
- Recognized for outstanding academic presentation skills in consecutive years
- Zhiyuan Honor Scholarship, Shanghai Jiao Tong University (2019-2023)
- Consecutive four-year scholarship for students in the Zhiyuan Honors Program
- Zhiyuan’s Leadership Scholarship, Shanghai Jiao Tong University (2022)
- For students demonstrating exceptional leadership and academic excellence
- National First Prize, RoboCup Robot World Cup China Competition (2022)
- Led team to victory in China’s premier robotics competition
- Outstanding Winner, Mathematical Contest in Modeling (MCM) (2021)
- Achieved the highest recognition in the international mathematical modeling competition
Research Interests & Future Goals
- Advancing cross-morphology learning for universal humanoid control systems
- Developing next-generation multimodal AI integrating vision, language, and physical reasoning
- Creating scalable frameworks for real-world deployment of intelligent humanoid systems
- Bridging the gap between laboratory research and industrial applications in humanoid robotics
- Contributing foundational technologies for humanoid robots in human-centric environments