competition DARPA TIAMAT Challenge — 1st Place Robust Spot policies for high-speed exploration, fall recovery, and obstacle avoidance in large-scale environments. Machine Common Sense DARPA-sponsored egocentric tool-use task — 2nd place vs. CMU, UC Berkeley, MIT, and IBM. explorations Go2 Locomotion via Adversarial Motion Priors Natural, dynamic quadruped gaits for Unitree Go2 trained with Adversarial Motion Priors. Whole-Body Control for Go2 + D1 Arm (Teleoperated) Teleoperated pick-and-place with a Unitree Go2 quadruped and D1 arm, showcasing whole-body controller robustness. Learning to Face a Chair from Egocentric Depth RL policy that orients a robot to face a chair using only an onboard depth camera, no external sensing. Learning to Push a Chair from Point Cloud Input RL policy that pushes a chair using onboard point cloud observations, no external sensing. class-project Language for Model-Based Reinforcement Learning Multi-task natural-language control of a robotic arm via seq2seq trajectory generation. Self-Balancing Platform via Reinforcement Learning Sim-to-real RL for a 2-DOF self-balancing platform, deployed on low-power hardware. CNN-RNN-Transformer Late-Fusion Network Transformer/CNN/RNN fusion model analyzing both audio and lyrics of songs. capstone Bipedal Obstacle Avoidance and Point-Goal Navigation (Cassie) Vision-augmented heuristic navigation layered on top of a blind RL locomotion controller. opensource IsaacLab Contributions to NVIDIA's open-source framework for unifying robotics research workflows. RSL-RL Contributions to a fast, GPU-native reinforcement learning library. MuJoCo Bug reports filed against the MuJoCo physics simulator. Weights & Biases Bug reports filed against the Weights & Biases experiment tracking toolkit.
