Emotionally Expressive Humanoid Robot

As Lead Mechanical Engineer at SDSU's JSB Center for Artificial Intelligence, I'm heading development of a humanoid robot capable of emotionally expressive, day-to-day task execution using machine learning.

The robot features anatomically correct arm and hand kinematics with 5 degrees of freedom, driven by 10 quasi-direct drive planetary gear actuators daisy-chained together. Autonomous navigation runs on a NVIDIA Jetson Thor, supported by LiDAR and ultrasonic sensors (STM32-controlled) for real-time obstacle detection. The chassis panels are fabricated from twill-weave carbon fiber using epoxy wet layup and vacuum bagging for optimized weight and structural performance.

Motion planning uses ROS2 and MoveIt 2, with Isaac Sim for simulation and kinematics validation. The robot runs entirely locally — no cloud computing — for a secure, private platform with an estimated 4–6 hour runtime under mild load. Current functionality includes crossing its arms, folding laundry, and reaching overhead to verify full range of motion.