New Release | NVIDIA AI Perception for ROS Developers

Category：technology
views：114

All moving things will become autonomous. All autonomous things require advanced real-time perception.


This latest announcement by NVIDIA, provides a complete set of perception technology to the ROS developer community. For developers seeking to incorporate advanced computer vision and AI/ML functions into ROS-based robotic applications, this will shorten development time and improve performance of . Brian Gerkey



Open Robotics will be extended for NVIDIA AI of ROS Open Robotics CEO


said: "As more and more ROS developers utilize hardware platforms, these platforms include additional computing functions designed to reduce the burden on the host CPU, ROS is being developed to make it easier and more effective to use these advanced hardware resources.Our cooperation with an accelerated computing leader like NVIDIA and its rich experience in artificial intelligence and robotics innovation will bring great benefits to the entire ROS community.


NVIDIA and Open Robotics have reached an agreement to accelerate ROS 2 performance on the NVIDIA Jetson edge AI platform and GPU-based systems, and implement Open Robotics 2 performance on Omniverse and Omniverse. Seamless simulation interoperability between NVIDIA Isaac Sim.


NVIDIA Jetson The platform is designed to provide high performance for robotics experts in many applications. Delayed processing makes it responsive, safe and collaborative. Open Robotics will upgrade ROS 2 to efficiently manage the data flow between the GPU and other processors on the NVIDIA Jetson platform and share memory, making it responsible for real-time processing The application performance of high-bandwidth data from sensors such as cameras and lidars has been greatly improved.


In addition to enhancing the deployment of robot applications on Jetson, Open Robotics and NVIDIA are developing to integrate Ignition Gazebo and NVIDIA Isaac Sim _strong163span s plan.NVIDIA Isaac Sim already supports ROS 1 and 2 out of the box, and creates a vital 3D content ecosystem by connecting popular applications such as Blender and Unreal Engine 4.


Ignition Gazebo has set many records in the entire robotics industry for decades, including the ongoing DARPA Subterranean Challenge and other high-profile competitions.


By connecting these two simulators, ROS developers can easily move their robots and environments between Ignition Gazebo and Isaac Sim to run large-scale simulations, and can also use high-fidelity dynamics and precise sensors The advanced functions of the two simulators, such as model and realistic rendering, are used to generate synthetic data for training and testing AI models.


The software resulting from this collaboration is expected to be released in the spring of 2022 by .


The Isaac GEM released for ROS is significantly faster


is a platform for developers to build on ROS that makes it easier to build a high-performance IROS span1sac on ROS. The hardware acceleration package of the solution. These GEMs focus on improving the throughput of image processing and DNN-based perception models, which are becoming increasingly important for robotics experts.These packages reduce the load on the host CPU while significantly improving performance.


new for Isaac GEM ROS includes:

  • SGM Stereo Disparity and Point Cloud
  • Color Space Conversion and Lens Distortion Correction
  • AprilTags Detection


Figure 1 ROS support for stereo cameras, you can see the left and right camera views in the Rviz tool of ROS. Rviz can display RGB and depth images.


Isaac Sim’s new features provide convenience to ROS developers


I1span, which includes the latest version of support for ROS developers. The focus is on the ROS2 navigation stack and the MoveIt motion planning framework.These examples are available today and can be found in the Isaac Sim documentation:


List of ROS examples in Isaac Sim

  • ROS Sparespan _span _
  • _span _span

    • ROS Span _span
    • ROS Navigation
    • ROS TurtleBot3 Sample
    • ROS Manipulation and Camera Sample
    • ROS Services
    • MoveIt Motion Planning Framework
    • Native Python ROS Usage
    • ROS2 Navigation


    Figure 2 Isaac Sim on Omniverse functional module diagram,Contains robot model, environment model and 3D data input.


    Isaac Sim generates a synthetic data set for training perception


    Isaac is also equipped with a set of powerful simulators. Synthetic data can be generated to train and test perceptual models. As robotics experts integrate more perception functions into their platforms, these functions will become more important. Obviously, the stronger the robot's ability to perceive the environment, the stronger its autonomy and the less human intervention required.


    Once the synthetic data sets are generated by Isaac Sim, they can be directly input into NVIDIA TAO (an AI model adaptive platform) to adjust the perception model according to the specific working environment of the robot. Before collecting any actual data from the target environment, it is possible to ensure the performance of the robot's perception stack in a specific work environment.


    For a long time, robotics experts have been facing various challenges in connecting and integrating classic robotic tasks (such as navigation) and AI perception stacks. Isaac Sim solves this workflow challenge by acting as both a robotic and synthetic data generation tool and integrating with the TAO training platform.


    For more exciting content, stay tuned for ROS World and GTC 2021


    strong on October 13th, 2021 ’s ROS World prepares.We plan to release more new GEMs for Jetson developers, including multiple commonly used DNNs. We will also release a number of Isaac Sim features that support the ROS developer community. Welcome to our virtual booth, participate in the NVIDIA ROS roundtable, and watch the technical demonstration of Isaac Sim and other exciting content.


    In the upcoming GTC conference will be held from November 8th to 11th , NVIDIA will bring you wonderful lectures and lectures. We have prepared a channel specifically for robot developers, including a speech by Brian Gerkey, CEO and co-founder of Open Robotics. In addition, we will also hold lectures covering topics such as NVIDIA Jetson, Isaac ROS, Isaac Sim, and Isaac GYM.


    leave immediately


    immediate experience NVIDIA AI Perception, please use the following links:


    download for ROS of Isaac GEM

    https://nvidia-ai-iot.github.io/ros2_jetson/ros2-packages/

    https://nvidia-ai-iot.github.io/ros2_jetson/ros2-packages/

    https://nvidia-ai-iot.github.io/ros2_jetson/ros2-packages/ Get / isaac-sim

    tutorial on synthetic data generated using the Isaac Sim

    https://www.youtube.com/playlist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X

    Learn more about accelerating machine learning training with the Tao toolkit

    https://developer.nvidia.com/tao


    Webinar dry goods are coming

    This webinar will introduce the robot simulation platform Isaac sim in depth.Including model import, communication with ROS, navigation and perception AI part based on simulation environment development, as well as introduction of GPU-accelerated Isaac ROS algorithm part.



    strong13 strong registration is now open! will be an online event from November 8 to 11, bringing together thousands of industry developers, AI innovators, and business leaders. Welcome to scan the QR code of the poster below, join GTC, and join in the feast of science and technology!


    Huang Renxun keynote speech will be broadcast on November 9th (Tuesday) at 4 pm, so stay tuned for more exciting content!


    *This article is transferred from NVIDIA Enterprise Solutions

    .
technology Latest News
Site article recommendation
technology Video
↑