GR-00 / GR0X Morphing Humanoid

Problem

Robots are usually built around one body type, and a platform tuned for one task is a compromise for every other. GR0X asks a different question: can a single humanoid reconfigure its own morphology to take on the advantages of multiple robot body types — changing limb length to trade off strength, reach and speed for the task at hand?

My role

I founded GR0X and serve as its research lead in the ALMaS Research Group at WPI, leading a 12-person research team (including 3 M.S. students and 5 seniors) through WPI’s Major Qualifying Project program. I architected the control system, direct the mechanical and electrical integration across the platform, and advise 2 of the lab’s 3 M.S. thesis projects, guiding research on morphology–performance trade-offs (strength, reach, speed).

I also secured sponsorship and in-kind support from Microchip Technology, Polymaker and G20 Inc., funding hardware, materials and compute.

Approach

• Adaptive lower body. A planetary lead-screw prismatic limb mechanism lets each leg extend from roughly 71 cm to 100 cm. A provisional USPTO patent has been filed.
• Distributed multi-brain control (ROS 2). A Jetson AGX Thor runs the reinforcement-learning locomotion policy and talks over ROS 2 on Ethernet to a Raspberry Pi 5, which runs a 200 Hz motor control loop across two CAN FD buses (see robstride-motor-control).
• Actuation. RobStride RS-02, RS-03 and RS-04 actuators.
• Power. A 48 V 13S4P battery with hierarchical power distribution.
• Learning. Locomotion policies are trained in Isaac Lab (see RL Locomotion Training).

Results

• Adaptive lower-body / leg subsystem fully validated, with working RL-based locomotion policies.
• Provisional patent filed; mechanism submitted to the 2026 ASME Student Mechanism and Robot Design Competition.
• In progress: the upper-body subsystem (torso, arms, actuation) and preparation for sim-to-real transfer.