The Fascinating Future of AI-Based Robotics

Whether you have a Roomba vacuum scouring your home for dirt and debris, or you’re lucky enough to have experienced some level of self-driving, most people realize that advanced robotic applications rely on intelligent artificial vision to navigate the terrain and avoid collisions. However, the truly intelligent robots of the future will not only need eyes to see and navigate, but also intelligence to interact, communicate and make decisions on their own. With pre-trained models and even live training, many robots of the future will need to infer and ingest important information from the world around them, interact seamlessly with humans, and also make the right decisions and recommendations autonomously. This area of ​​smart machines is a huge market opportunity for which NVIDIA has been developing myriad solutions with its Jetson portfolio for years now.

From the little one, $59 Jetson Nano development kit for manufacturers (2019), to its latest complete Jetson AGX Orin development kit which sells for $1999, and is a derivative of the company’s Drive AGX Orin for self-driving vehicles, NVIDIA has focused not only on developing the silicon engines that can power the next generation of autonomous machines and smart edge devices, but also a complete software stack and deployment platform that developers can leverage.

Launch Tires on Next-Gen AI for Robots and Intelligent Edge

NVIDIA’s Jetson AGX Orin is available both in a development kit with 32 GB of onboard LPDDR5 memory, as well as in a family of modules for various embedded applications. My team had hands-on time with the SDK, and the brains of the operation is NVIDIA’s Jetson AGX Orin processor in its full implementation, with a complex of 64-bit 12-core Arm Cortex-A78AE processors that straddles a shotgun alongside an NVIDIA GPU Ampere which sports 2,048 CUDA cores and 64 Tensor cores. . Think of the performance of the GeForce RTX 3050 class, but leveraged here for machine learning, inferring workloads specifically, but also with some training capabilities, which I’ll cover a bit.

Jetson AGX Orin’s high-end configuration offers 275 TOPS (Trillion Operations Per Second) of computing power, although Orin cards can scale from 15 to 60 Watts, to meet the power needs of various low-power peak applications . The development kit also brings tons of additional I/O for access and adaptability, including the usual JTAG port and 40-pin GPIO (General Purpose I/O) header, in -Additional sensor heads and enough USB ports to connect not just cameras and mics, but peripherals and external storage. If that wasn’t enough, there’s also a 10 Gigabit Ethernet port and a pair of PCI Express slots. In a nutshell, NVIDIA’s Jetson AGX Orin SDK is designed to provide unrestricted access to as many platform features and capabilities as possible.

Advanced AI hardware developers need powerful software tools and infrastructure

NVIDIA bundles Jetson AGX Orin with Ubuntu Linux and its Jetson Jetpack SDK. Jetpack is NVIDIA’s development environment. Jetpack works in conjunction with NVIDIA ROS (Real-Time Operating System), pre-trained models, and various other SDKs and libraries to support a myriad of applications such as machine vision, real-time speech, and speech. ‘analysis. In short, the Jetson Software suite of tools is designed to enable engineers and developers from a myriad of disciplines, from robotics to industrial automation, IoT and more, to program and develop in their language. kindergarten while optimizing and targeting the Jetson platform. To give you an idea of ​​what can be done with these tools and Jetson AGX Orin, Ben Funk on hot hardware was able to quickly tap into NVIDIA’s Riva AI voice SDK for a fun real-time demonstration. Check it out…

As you can see, NVIDIA’s Riva AI speech recognition engine has no problem keeping up with Ben’s cadence, understanding his pauses and minor corrections, and ultimately understanding conversational context for better accuracy. Pair Riva Automated Speech Recognition (ASR) with Conversational AI and you have a chatbot that you can customize with additional NVIDIA tools and SDKs, as the company itself did in this demo…

NVIDIA’s Orion bot is fully powered by Jetson AGX Orin and demonstrates machine vision and location recognition for its navigation, conversational AI, recommendation AI, and voice recognition. Nvidia Omniverse Avatar then brings all of these technologies together for a cute and useful traveling digital construction assistant, again, all powered by Jetson AGX Orin.

Beyond the trained vision models provided by NVIDIA for Jetson AGX Orin, the company also provides its TAO (Train, Adapt, Optimize) toolkit which gives developers easy access to adapt the models to their own applications and create additional features locally, or deploy models to NVIDIA’s DeepStream SDK in the cloud. deep stream can run on NVIDIA A100-powered data center servers, on Microsoft Azure Cloud infrastructure, for a major increase in performance and compute resources, although it can also run locally on-premises and at the edge on Jetson itself -same.

NVIDIA Jetson AGX Orin Development Kit Leverages Fully Intelligent AI Edge Platform Solution

As a technology analyst, for me the key takeaway here is that NVIDIA’s Jetson AGX Orin Development Kit is a powerful suite of silicon engine tools and software for robotics, autonomous machines, and other AI edge IoT (Internet of Things) devices. However, the GPU giant has taken the approach of enabling a complete platform solution, with trained models, the ability to continuously adapt and train them, as well as SDKs for machine vision, speech, conversational AI and more, plus the ability to deploy and tune locally or in the cloud.

I believe that this true platform solution approach is the only way to really make AI or any type of IT solution adaptable properly. Competing solutions such as FPGA-based platforms from Xilinx (AMD) take similar holistic approaches, but NVIDIA has made a significant leap in this market, having pioneered the acceleration and development of AI on GPUs for over a decade now, when its CUDA programming language was first introduced. back in 2007.

NVIDIA currently offers Jetson AGX Orin development kits for $1999 through its network of distributors. Various configurations of the Jetson AGX Orin production modules above will begin shipping in July, with lower cost Jetson Orin NX modules shipping in September. However, I expect NVIDIA channel partners to already turn on more than a few orders for these early Jetson AGX Orin dev kits.

Leave a Comment