NVIDIA’s Isaac SDK brings modern AI to autonomous machines

Robotics Summit & Expo 2019 logo

SAN JOSE, Calif. — Robotics developers are off to the races creating autonomous machines of the future, said NVIDIA Corp. this week. The company said it’s fueling their efforts with the NVIDIA Isaac software development kit, which will be publicly available soon as a free robotics developer toolbox for accelerating innovation and deployment.

NVIDIA founder and CEO Jensen Huang announced the news to a packed crowd at the annual GPU Technology Conference (GTC) here.

Building robots has historically been difficult, said the processor maker. Delivering robot intelligence has been deterred by a lack of unified and easy-to-use software and hardware platforms. Robotics development at scale has, until recently, required custom programming to simulate robots in operation.

The Isaac SDK robotics developer toolbox is designed to change that. It moves to general availability in April.

This toolbox offers developers access to Isaac applications, GEMs (robot capabilities), a Robot Engine, and Sim. Huang claimed it will save manufacturers, researchers, and startups hundreds of hours by making it easier to add AI for perception, navigation, and manipulation into next-generation robots.

At the core is Isaac Robot Engine, which enables developers to build modular and high-speed robotics applications easily deployed to Jetson platforms, including NVIDIA‘s robot reference designs.

There are also more than 20 GEMs, providing plug-and-go functions for hard robotics problems such as navigation and perception algorithms. GEMs are robotics algorithm modules that are built and optimized for the NVIDIA Jetson AGX platform.

What the industry has previously lacked was a complete platform built on an ecosystem of trusted industry-standard software and hardware. To help achieve that, NVIDIA is making its robot hardware reference designs open source.

Welcome back, Carter

Developers can use the Isaac SDK robot developer toolbox to create applications with NVIDIA’s Carter autonomous delivery and logistics robot, based on Jetson AGX Xavier.

Businesses of all sizes can use this to build applications that enable robots to navigate and move things about.

NVIDIA's Carter robot

Kaya bunga, Nano!

Researchers and casual makers alike can use the developer toolbox to create applications with Kaya, a robot based on our low-cost Jetson Nano, to create autonomous machines.

Kaya is built on NVIDIA's Jetson Nano

Jetson Nano

Jetson Nano comes in two versions: The $99 developer kit is ideal for makers, developers, learners, and students. The $129 production-ready module is for mass market AI products needing horsepower at the edge.

Low cost doesn’t have to mean low flops. Look under the hood: 472 GFLOPs of compute performance to run AI workloads.

The Jetson Nano Developer Kit is available now. The Jetson Nano module will be available in June 2019.

Isaac GEMs

GEMs are essential and fundamental robotics algorithm modules that are built and optimized for the NVIDIA Jetson AGX platform.

For example, GEMs enable developers to easily add SLAM navigation features — obstacle avoidance, lidar-based localization, path planning, and map creation. They can also control interactions with humans such as voice command detection.

Isaac Robot Engine

Isaac Robot Engine offers a framework that allows developers to easily write modular applications for deployment on robots. Isaac Robot Engine is optimized for performance across the entire NVIDIA Jetson family — Jetson AGX Xavier, Jetson TX2 and Jetson Nano along with GPU-powered workstations.

Isaac Robot Engine delivers to users advanced monitoring and debugging capabilities required to build complex and highly parallel compute graphs. Capabilities include logging and replay of sensor data, efficient memory management and live, Web-based visualization.

NVIDIA's Isaac robot engine

Isaac Sim

Isaac Sim allows developers to test robots — without breaking things — in simulation, said NVIDIA. Once tested, applications can be deployed on robots running on Jetson.

It’s easy to do this with the Isaac SDK, allowing for high-fidelity simulation and advanced real-time rendering. Isaac Sim is tightly coupled to the tools and framework in this toolbox, making for easy transfer of data and algorithms with robots — key for accelerating robot development under deadline pressures.

Isaac Sim is a virtual robotics laboratory. It’s a rich 3D world simulator that enables research and design for robots of the future at lower cost and time than would be possible without simulation, said NVIDIA.

Soon robots will be everywhere, and many people will work and live alongside them. Safety demands new and better ways to design and test robotics systems.

With Isaac Sim, engineering and testing that could take months can happen in minutes.

Start revving your robotic engines

In April, the Isaac SDK will generally available for download. Additional details on reference designs will be shared soon.

Robotics Summit & Expo 2019 logoKeynotes | Speakers | Exhibitors | Register

Also at GTC, Liqid announced a 20-GPU chassis that will support the NVIDIA T4, V100 Tensor Cores, and Quadro RTX processors. Iguazio, whose high-performance platform supports serverless and machine learning applications, unveiled native integration with NVIDIA GPUs. Storage provider DataDirect Networks plans to showcase its AI reference architectures for autonomous vehicles and other applications.

Editor’s note: This post by Claire Delaunay, vice president of engineering at NVIDIA, was originally published by NVIDIA.

The post NVIDIA’s Isaac SDK brings modern AI to autonomous machines appeared first on The Robot Report.