Mitsubishi R&D

1 / 3

Unitree Go2 Robot with one of the LiDARs I tested.

2 / 3

Skiing trip to Niseko with my teammates:my friends throughout my internship!

3 / 3

A trip to the Osaka Mint sakura viewing with my coworkers!

❮ ❯

From February to May of 2024 I lived in Amagasaki, near Osaka, Japan, to intern at the Mitsubishi R&D robotics department. During my internship at Mitsubishi, I worked on enabling autonomous navigation for a Unitree Go2 quadruped robot by implementing SLAM (Simultaneous Localization and Mapping) to support tasks such as search and rescue, stair traversal, and indoor food delivery. I integrated and tested three different LiDAR sensors—Hesai XT16, Livox Mid360, and the Go2’s proprietary LiDAR—leveraging open-source GitHub repositories to build a flexible and modular SLAM pipeline.

I conducted a comparative study of SLAM algorithms, evaluating them based on motion modeling, computational efficiency, and mapping accuracy. I tested map generation in both open lab environments and cluttered office spaces with dynamic obstacles to determine which algorithm performed best under varying conditions. This analysis informed our selection of the most robust SLAM solution for real-world deployment.

To support the hardware integration, I designed and 3D-printed custom mounts for the LiDAR sensors and an additional NVIDIA Jetson module required for onboard computation. I also configured auxiliary power supplies and routing to ensure stable operation during high-mobility tasks. My work will enable the robot to autonomously navigate complex indoor environments, laying the groundwork for future applications in logistics and emergency response.