Robot manipulation and locomotion strategy development:

• Develop intelligent control policies for bimanual manipulation and whole-body locomotion based on Reinforcement Learning (RL), Imitation Learning (IL), and Diffusion Policy methods. Enable whole-body coordination (Whole-body Coordination) for complex scenarios such as object carrying, tool use, walking, stair climbing, and obstacle traversal.

Simulation training and sim-to-real deployment:

• Conduct large-scale policy training and evaluation on high-performance physics simulation platforms (e.g., Isaac Lab, MuJoCo). Research sim-to-real transfer methods and deploy trained policies onto real humanoid robots.

Algorithm innovation and performance optimization:

• Research training methods for multi-task, multi-modal, and multi-robot shared policies to improve generalization and stability. Integrate vision–language–action models (VLM / VLA) to achieve task understanding, decision-making, and closed-loop control.

‍

Requirements:

• Major in Computer Science, Artificial Intelligence, Mathematics, Robotics, or related fields;
• Strong coding skills and solid algorithmic foundations with excellent problem analysis and solving abilities.
• Solid understanding of robot kinematics/dynamics; familiarity with robot reinforcement learning techniques; hands-on experience with real-world implementation and optimization.
• Proficient in deep reinforcement learning (e.g., PPO, SAC, TD3, Dreamer, GAIL) or imitation learning methods; practical experience or deep understanding of diffusion-based policies (Diffusion Policy / Guided Diffusion / Flow Matching) for robot control.
• Knowledge of trajectory optimization or control for high–degree-of-freedom robots; experience in motion control or navigation-related R&D.
• Proficient in C++/Python and Linux operating systems; strong interest in robotics and research enthusiasm.

‍