> CURRENT_TASK: Master's_Robotics_UIUC
> FOCUS: Humanoids, Safe Autonomy, Embedded AI
> STATUS: Developing Intelligent Systems
Stanley Controller | Pedestrian Prediction | ROS2 | YOLO
Developing pedestrian-behavior-based control policies for the GEM autonomous vehicle. Focusing on predictive braking and path planning in high-uncertainty environments using behavioral models.
HRI | Motion Tracking | Real-time Control
Building a closed-loop system where a robot tracks a human hand to dynamically hand over objects. Instead of the human reaching out, the robot adapts to the human's position for a seamless, stable transfer.
PyTorch | Model-Free Algos | RGB-D
Implementing and comparing model-based vs. model-free algorithms for estimating the 6D pose (position + orientation) of texture-less industrial objects in cluttered scenes.
Trained ANYmal quadruped policies using Proximal Policy Optimization. Achieved robust traversal on irregular terrain with 0-fall safety constraints.
Fused Visual-Inertial Odometry with footstep planning. Enabled autonomous navigation in GPS-denied environments with <10cm drift.
Implemented whole-body motion planning for HRP-4. Enforced ZMP constraints for stable manipulation and reaching tasks.
Developed a perception pipeline converting depth clouds to elevation maps for quadruped footstep placement.
Built a 5-DOF prosthetic arm controlled by EMG signals. Used XGBoost on ESP32 for real-time gesture classification.
Integrated CBF as a safety filter for RL policies to guarantee stability during training.
> # THE OBJECTIVE
The robotics industry is building calculators. I want to build the smartphone.
> # THE CONCEPT
"One Robot. Many Lives."
My vision is a Universal Chassis augmented by a Skill Ecosystem. Instead of single-purpose hardware, we need a core intelligence that can download "Physical Apps" - mastered skills for logistics, caregiving, or disaster response.
> # THE IMPLEMENTATION
The key is not just mechanics, but the Core Intelligence - an OS that translates human intent into flawless execution using verified skill blocks. This system must be built on a foundation of "Simulation-First" safety protocols.