Job Description
The research project focuses on developing real2sim2real learning and control methodologies for acquiring robust dexterous manipulation skills in complex, content-rich environments.
The research engineer is expected to contribute to the following key aspects:
a) Design simulation-based skill learning and optimization methods for contact-rich dexterous manipulation, integrating differentiable simulation, trajectory optimization, reinforcement learning, and/or imitation learning to efficiently acquire transferable manipulation policies under complex physical constraints and diverse task conditions.
b) Develop sim-to-real adaptation and closed-loop skill refinement approaches that leverage multimodal sensory feedback, including vision, tactile, and proprioceptive information, to bridge the simulation-to-reality gap and continuously improve manipulation performance in the real world.
Job Requirements
- Master or Bachelor Degree in a relevant discipline, e.g. robotics, computer science, electrical/electronic engineering, etc.
- Extensive experience in dexterous hand manipulation, especially in in-hand skills with demonstrated expertise in real-world robotic applications.
- Substantial practical experience in real robot development and deployment, including multi-fingered robot systems.
- Proficient experience with various robotics simulators (e.g., MuJoCo, ISSAC) and excellent programming skills in C++ and Python with the ability to develop efficient, production-quality code.
- Open to Fixed Term Contract.