Developer Blog: Building an Intelligent Robot Dog with the NVIDIA Isaac SDK

The modular and easy-to-use navigation stack of the NVIDIA Isaac SDK continues to accelerate the development of various mobile robots. Isaac SDK 2020.1 includes support for quadruped robots and adds to the other robot platforms supported in the SDK, such as differential wheel bases and holonomic wheel bases.

Laikago, a quadruped robot by Unitree Robotics, is the newest member of the mobile robot portfolio. Laikago is powered by NVIDIA Jetson TX2 and features a low-level controller for balancing and walking that allows remote controlled locomotion. This post explains how you can grant autonomous navigation to Laikago, even if you’ve never used the Isaac SDK before.

Out of the box, Laikago lacks any sensor that can perceive the surroundings. By default, the robot is not capable of localization or obstacle avoidance. To obtain 3D information about the environment, we installed a Velodyne VLP-16 LIDAR on Laikago for this project. You can download 3D-printable mesh files used for mounting the LIDAR, which are based on the CAD Model of Laikago. We also installed a DC-DC converter to decrease the 19V output from Laikago and meet the voltage range of the LIDAR. A USB to Ethernet adapter connects the LIDAR to the NVIDIA Jetson.

Read the full tutorial, Building an Intelligent Robot Dog with the NVIDIA Isaac SDK, on the NVIDIA Developer Blog.