During the GTC 2022 conference, Nvidia introduced new hardware and supporting services for the field of robot development, aiming to help industry development and machine testing. First, Nvidia's Isaac Sim robot simulation platform will soon provide cloud access. Second, the lineup of system-level modules is also rapidly expanding, including the Jetson Orin Nano designed specifically for low-power robots, and a new platform called IGX.
As early as June last year, Isaac Sim was launched in public beta. features allow designers to simulate robots interacting with real-world models, such as digital reconstructions of warehouses and factory workshops.
users can use simulation sensors to generate data sets, train models on real-world robots, and use batch parallelism and unique simulation synthesis data to improve the performance of the model. The expressions like
are not gimmicks at the marketing level. Some studies have pointed out that synthetic data can help companies that try to implement AI solutions but encounter many development challenges to bail. Researchers at
MIT (MIThings 7) have also recently found a way to classify images using synthetic data.
What's more, major autonomous vehicle companies are actively using simulated data to fill their shortcomings in collecting real data from the road.
Nvidia says the upcoming Isaac Sim is available on AWS RoboMaker and NVIDIA NGC and supports deployment to any public cloud and will soon be available to the company's Omniverse Cloud platform.
Enterprise customers will be able to use real-time fleet task assignment and route planning engines on the cloud platform and use NVIDIA cuOpt to optimize the robot's path planning performance.
Gerard Andrews, senior product marketing manager at Nvidia, wrote in a blog post:
Cloud Isaac Sim can bring together development teams distributed around the world to share a virtual world where they simulate and train robots.
Meanwhile, Isaac Sim running in the cloud means developers no longer rely on powerful workstations to run simulations and can configure, manage, and view simulation results on any device.
Next talk about Jetson Orin Nano: Nvidia launched Jetson Orin in March, the company's next-generation ARM single-board machine for edge computing .
The first member of the series is Jetson AGX Orin, but we are now welcoming Orin Nano – which expands the Jetson Orin family’s product lineup with a more affordable configuration.
Orin Nano has the smallest form factor of the series to date, and can perform up to 40 trillion operations per second (TOPS) at full speed.
As the most entry-level SKU for the Jetson family, Nvidia also offers six Orin-based production modules—for a range of robots and local offline computing applications.
It is worth mentioning that Orin Nano adopts modules compatible with Nvidia's previously announced Orin NX to support the AI application pipeline with the Ampere GPU architecture that was launched in 2020.
Interested developers can buy two versions in January next year, which will be priced at $199. Among them, the Orin Nano 8GB has configurable power of 7~15W and the performance is 40 TOPS; while the Orin Nano 4GB has lower power of 5~10W and the performance is 20 TOPS.
NVIDIA Vice President of Embedded and Edge Computing Deepu Talla said in a statement:
has received adoption from more than 1,000 customers and 150 partners since the launch of Jetson AGX Orin six months ago. And with the arrival of Orin Nano, this trend will expand significantly.
Also, when it comes to price, the Jetson AGX Orin costs well over $1,000, while the Orin Nano is much more affordable – it sets new standards for entry-level edge AI and robotics.
Finally, Nvidia quietly released a glimpse of IGX —As a set of platforms suitable for "high-precision" edge AI (especially for application scenarios such as manufacturing and logistics), the company said it can provide an additional layer of security and low-latency AI performance experience for highly regulated environments such as factories, warehouses, clinics, and hospitals.
As part of the IGX platform, IGX Orin is an AI chip used to fund industrial machines and medical devices.
Nvidia said that the development kit for prototypes and product testing will be provided to enterprises early next year.
Each toolkit has an integrated GPU/GPU, as well as a software stack with security features that can be configured and programmed for different use cases.
Nvidia added that it is currently working with developers of Canonical distributions such as Red Hat and SUSE to provide long-term and full-stack support for IGX. CEO Huang Renxun said in a statement:
As humans increasingly cooperate with robots, various industries are developing new functional safety standards for artificial intelligence and computing.
and IGX will help companies build next-generation software-defined industrial and medical devices to enable them to operate safely in the same environment as humans.
