- A recent paper on quadrupedal navigation in complex terrains was reported by IEEE Spectrum.
- Proposed MARG, a framework for perceiving the environment and planning footholds to avoid falling into gaps.
- Developed an imitation learning framework that can imitate aggressive motions from monocular videos.
- Created ConsJump, a hierarchical learning framework enabling quadruped robots to transition between walking and jumping over constrained spaces.
- Designed a unified policy containing a perception module and a locomotion module for recognizing different obstacles and traversing them using diverse skills.
- Invented TumblerNet, allowing quadrupedal robots to perform bipedal locomotion in challenging terrains and recover quickly after disturbances.
Research Experience
Research focuses on various control techniques to enhance the autonomy of robotics, particularly in quadrupedal navigation, perception, and locomotion strategies across complex terrains.