Today, NVIDIA announced the release of ISAAC SDK 2019.2, bringing numerous features and enhancements to the robotics community.
The NVIDIA® Isaac Software Development Kit (SDK) is a comprehensive toolbox for accelerating the development and deployment of AI-powered robots.
With the latest release, we are enabling full capabilities for AI perception and navigation making it easier to build next-generation robots. This means roboticists, developers, researchers and students can now easily get started in learning and building unique projects powered by Isaac SDK.
Release Highlights:
Robot Engine
Subgraphs
- Isaac Robot Engine uses computational graphs and entity component system to help developers to break down a complex robotics use case into small building blocks. With subgraphs, developers can easily re-use existing node graphs. One of the main benefits include making the software modular and dramatically reducing the number of lines of code.
Distributed Workspaces
- Isaac SDK uses Bazel for building/testing. While building, Bazel can pull in external dependencies, typically source code located outside main repository as long as these workspaces follow certain rules.
- Isaac Distributed Workspace on github provides example(s) for partners (such as sensors, algorithms, used by Isaac SDK) to create such compatible workspaces and contribute directly to Isaac SDK.
- The GitHub repository hosted by NVIDIA Isaac Robotics team, provides a mechanism to track dependencies between packages, and separate out dependency-heavy packages into separate workspaces.
GEMs (Robot Skills)
Navigation Stack Updates
- Enhancements for differential/holonomic dynamics for robots that are not ‘circular’.
- A new customizable Robot shape model based on a union of circles i.e. custom shapes can be approximated (for e.g. Rectangles with 8 circles and Triangles with 7 circles etc.).
- Waypoint and trajectory planner updated to work with new union of circle model.
Superpixels
- Superpixel algorithms are pre-processing algorithms to help with object segmentation, object tracking etc. This new feature accelerated by GPU adds to the perception capabilities for Isaac based robots to avoid small obstacles during navigation.
Simulation
- Get to visualize and build scenarios without the need of in-house 3D expertise with Unity 3D Game Engine support.
Availability
Members of the NVIDIA Developer Program can get free access to download Isaac SDK.
