Humanoid Locomotion as Token Prediction

RL for Versatile, Dynamic, and Robust Bipedal Locomotion Control

Learning Torque Control for Quadrupedal Locomotion

Real-World Humanoid Locomotion with Reinforcement Learning

Robust and Versatile Bipedal Jumping Control through Multi-Task Reinforcement Learning

Dynamic Quadrupedal Robotic Goalkeeper

Hierarchiral RL for Precise Soccer Shooting

Adapting Rapid Motor Adaptation for Walking on Varying Terrain

Collaborative Navigation of a Cable-towed Load by Multiple Quadrupeds

Dynamic Quadrupedal Locomotion over Stepping Stones

Bipedal Navigation in Height-Constrianed Environments

Autonomous Navigation for Quadrupedal Robots

Robotic Guide Dog

Ball Juggling on the Bipedal Robot Cassie

Feedback Control for Autonomous Riding of Hovershoes by a Cassie Bipedal Robot

Variation Based Extended Kalman Filter on S2

Dynamic Walking over Randomly Placed Stepping Stones on ATRIAS

Geometric Control for a Quadrotor with a Payload Suspended through an Elastic Cable

Decentralized Control for Dynamic and Stable Walking for Exoskeletons

Dynamic Bipedal Walking over Stepping Stones with Random Step Length and Step Width

Dynamic Walking over Moving Stepping Stones while subject to Model Uncertainty

Safety-Critical Control for Quadrotor Aerial Flight

Quadrotor UAVs Cooperatively Manipulating a Shared Cable-Suspended Payload

The Reaction Mass Biped - A Variable-Inertia Biped Model with Proof-Mass Actuators

Differential-Flatness and Geometric Control for Multiple Quadrotors with a shared Cable-Suspened Point-Mass Load

Almost-Global Geometric Control for a Quadrotor with a Cable-Suspended Load

Avian-Inspired High-Speed Aerial Grasping at up to 3 m/s

Dynamic and Agile Running Gait on MABEL - Top speed 3 m/s

Our research lies at the intersection of applied Nonlinear Control and Hybrid Dynamic Robotics. Our goal is to design controllers for achieving dynamic, fast, energy-efficient, and robust maneuvers on hybrid and underactuated systems such as legged and aerial robots. This will require addressing the challenges of high degree-of-freedom, high degree-of-underactuation, nonlinear and hybrid systems with unilateral constraints, operating in stochastic and hard-to-model regimes. Our work has focused on creating dynamic bipedal locomotion - both walking and running, and dynamic aerial manipulation maneuvers.