Accelerating ROS 2 With NVIDIA GPU-Powered Libraries and AI Models#

Simulation gives us powerful tools for developing robotics applications, to train and test varieties of environments and situations without needing physical robots.

Using GPU-accelerated tools within simulation helps this process happen faster, and supports complex tasks such as object detection and pose estimation. Finally, by integrating these tools within the popular robotics framework ROS, specifically ROS 2, we create a path for these simulations to perform just as well on real robots as they do in simulation.

In this module, we will introduce ROS 2, Isaac ROS, and NVIDIA reference workflows to develop robotics applications. We will then perform an obstacle-aware pick and place operation, performed by a robotic gripper in Isaac Sim.

Learning Objectives#

In this module, you will:

  • Learn to utilize NVIDIA’s GPU-powered libraries to accelerate ROS 2 workloads, enhancing AI and robotics performance in real-time applications

  • Gain experience in leveraging GPU-accelerated ROS 2 packages to optimize robotic workloads

  • Understand the fundamental Isaac ROS building blocks

  • Understand techniques for improving performance and reducing latency in ROS 2 packages through GPU acceleration

  • Build an end-to-end perception-driven pick-and-place workflow for a robotic manipulator

On this page