I am a fourth-year undergraduate student at the University of Science and Technology of China (USTC). Starting Fall 2025, I will be joining the Hybrid Robotics Group at UC Berkeley as a Ph.D. student under the supervision of Prof. Koushil Sreenath.

Research Interests

My interest lies in Control and Robot Learning. I am currently focused on bridging high-level intelligence with low-level controller in physical world.

Research Experience

Language-directed Humanoid Whole Body Control via End-to-end Learning

• An end-to-end method enabling humanoid robots to interpret and execute diverse whole-body motions directly from natural language commands.

Learning for Safe Multiple Heterogeneous Quadcopter Control via Generalizable Barrier Certificate

• Developed a heterogeneous safe controller for quadcopters with generalizable barrier certificate.
• Simultaneously learn the control barrier function and constrained control policy.
• Online system identification from observation history for heterogeneous adaptation.

Geometry Variation Model Predictive Control For Bipedal Robot

• Geometry representation of system dynamic, invariant to coordinate system and free from singularity.
• Variation-based linearization of system dynamic to ensure the discretized system is energy-conserving.
• Enable stable walking of a bipedal robot. Convex MPC based on geometry representation of system dynamics. Free of singularity and energy preserving. 

State Estimation and Control of a Wheeled-Bipedal Robot

• Developed a robust state estimation and high-performance control algorithm for a wheeled-bipedal robot.
• Detect slipping based on the velocity difference of wheels’ contact points with the ground in the inertial frame to adjust covariance in state estimation Kalman filter.
• Designed LQR controller based on whole-body dynamics.
• More details can be found HERE

SLAM and Navigation of an Omnidirectional Robot

• Develop a navigation system for a mobile robot to navigate through uneven terrain.
• Enhance localization frequency and accuracy by integrating lidar, IMU, and encoder data using a Kalman filter.
• Analyze terrain traversability through point cloud segmentation.
• More details can be found HERE