The NVIDIA Isaac Software Development Kit (SDK) enables accelerated AI robot development workflows. Stacked with new tools and application support, Isaac SDK 2020.1 is an end-to-end solution supporting each step of robot fleet deployment, from design collaboration and training to the ongoing maintenance of AI applications.
Isaac SDK 2020.1, the latest version of Isaac SDK packs the tools and methodologies needed to develop, train, and test the robots in simulation using Isaac Sim on the collaborative virtual environment tool, Omniverse. It also includes all the proven features that helped BMW Group to develop their logistics robots.
NVIDIA recently announced that BMW Group implemented AI-enabled logistics robots developed using a unified software architecture, running on the NVIDIA Isaac robotics software platform. In this talk at GTC 2020, the BMW team demonstrates how they use the Isaac SDK to develop their STR (Smart Transportation Robot).
The Isaac SDK 2020.1 is an end-to-end solution that brings autonomy to many industries including manufacturing, healthcare, agriculture, and more:
- An open AI platform for robotics that’s built for customizations.
- A number of GPU-accelerated algorithms and Deep Neural Network (DNN) models.
- Accelerated development cycles using Isaac Sim to generate training data for DNN models & hardware in the loop testing.
Isaac SDK 2020.1 packs many new capabilities for manipulation, navigation, planning, and localization:
- 3D Object Pose Estimation: Accurate DNN models consist of object detection, 3D pose estimation, and pose refinement using depth sensor data. This can be used for various purposes e.g., for detection/pose estimation of a dolly in a warehouse for a robot to drive under or for detection/pose estimation of a container for a cobot (like Universal Robots’ UR10) to pick and place objects.
- Multi Lidar Support: In a grid-map based global localization and for particle-filter in continuous localization, having two lidars helps with robust localization in an ambiguous environment; and, reduced location uncertainty. Multiple lidars also help with much better perception (360 degrees) for collision avoidance.
- Path Planning with Costmaps for Navigation: This release offers a way to customize graphs, in an open space (like a factory) where robots can use optimal paths (i.e. have a notion of lanes) rather than taking the shortest path to the destination.
- Motion Planning for Manipulation: LQR-based motion planning is included for manipulation robots.
- Virtual Factory Space: A massive 125 meters x 50 meters area in Isaac Sim to train and test robots.
- Docker Container: Enables robotics ML engineers use Isaac Sim to generate synthetic images and train an object detection DNN (included with Isaac SDK) with NVIDIA’s Transfer Learning Toolkit.
- Experimental RL Application: Teaches a robot to navigate under a cart placed in front of it, using a deep neural network.
- Support for Quadruped: With support for quadruped robots like Unitree Laikago, this release adds to the other robot platforms supported such as differential wheel base, holonomic wheel base and 6 DoF Cobots.
- Support for New Sensors and Robots: This release also expands on the ecosystem of sensor and robot platforms that have embraced Isaac SDK. This includes support from our ecosystem partners including LIPS AE400 stereo camera, SICK Microscan3 Pro lidars, Basler cameras, Universal Robots’ Cobots, and more.
Isaac SDK 2020.1 represents a big step towards closing the sim-to-real gap, opening up new possibilities to build autonomous robots. Much of the robotics development – from creating a robot’s digital twin, developing robot applications to training/testing the robots – can be accomplished from the comfort of your home.
NVIDIA Isaac Sim 2020.1 is powered by Omniverse Kit for robotics and available to download today for anyone who signs up for the Early Access Program. Roboticists can take advantage of the virtual environment and reference applications to test faster, train better, and springboard into building complex robots with AI skills in navigation, planning, localization, and manipulation.
We welcome you to try out Isaac SDK 2020.1 and Isaac Sim 2020.1 today. Download them here.
Visit https://developer.nvidia.com/isaac-sdk to learn more about the Isaac Robotics Platform. Comprehensive SDK documentation can be found at https://docs.nvidia.com/isaac/isaac/doc/index.html. You can learn more about the Docker Container and try it out on NVIDIA’s NGC (the hub for GPU-optimized software).
We welcome your feedback! Be sure to visit our developer forum to share your experiences, questions, comments and to interact with other Isaac SDK developers.