I build robots that walk, map, and decide where to go next.
Growing up, Iron Man made me want to build machines. Now I build my own.
I'm Tanmay, a robotics engineer from India, studying for my Master's in Mobile Robotics at Universität Bonn. My work sits where a robot's body meets its sense of the world: getting legged machines to balance and walk, and getting mobile robots to explore and map places they have never seen.
The robot above is GS-01. It isn't a real machine yet. It is assembled from what I have actually worked on: the head from my SLAM research, the torso from whole-body control on a Unitree G1 humanoid, the legs from the Spot Micro quadruped I built and am now rebuilding in ROS2. The arms are still on the workbench.
The name Gaussian Space comes from the bell curve on its chest. Every robot lives in uncertainty, and the Gaussian is how it reasons about it.
I'm always happy to talk about legged robots, SLAM, or a role where I can work on either.
tanmayghai.de@gmail.comResearch, locomotion and hardware. Each one maps to a part of GS-01 on the home page.
Frontier-based active exploration with cluster-based zone selection and blacklisting. The robot navigates unknown environments on its own, building dense 3D maps while it searches for target objects. Implemented and debugged in ROS2 as supervised research at Universität Bonn.
Full-stack whole-body controller for the Unitree G1 humanoid. Stable standing balance, LIPM walking trajectory generation and bilateral sinusoidal arm motion, all visualised in MuJoCo. The next phase replaces the offline plan with online MPC (Herdt et al.) for reactive walking.
Built on the open-source Spot Micro project during my B.Tech. Assembled the 3D-printed hardware, wrote custom Arduino code for servo control and implemented keyboard teleoperation over WiFi. Integrated an MLX90640 thermal camera and YOLOv8 for real-time survivor detection in low-visibility environments.
What I'm building next.
Extending the G1 whole-body control stack to the Unitree Go2 quadruped. TSID-based loco-manipulation with Pinocchio and MuJoCo simulation, with a path toward autonomous navigation and real hardware.
Rebuilding the Spot Micro software from scratch on the same printed hardware: Raspberry Pi with a PCA9685 servo driver, a Pinocchio and TSID control stack, and Nav2 with frontier exploration. First milestone: it walks and explores on its own. The stack will be open source.
Education, projects and skills on one page.