Categories: Tech Explained

Deploying NVIDIA Jetson for AI-powered Automation

NVIDIA® Jetson™ has emerged as an early leader in the ongoing race for hardware platform supremacy to support the massive growth of artificial intelligence (AI) deployments. NVIDIA is already a household name in the world of advanced GPU technology, and their family of Jetson embedded SOMs (System on Module) combine ARM CPUs with specialized GPUs specifically optimized for matrix multiplication, a core component of most edge AI workloads.

In this blog post, we’ll delve into the value that NVIDIA Jetson offers for those deploying AI solutions, touch on the key differences between the various versions of Jetson, and lay out when and where Jetson is best leveraged for industrial automation applications.

Jetson vs discrete GPU and CPU

One of the key advantages of NVIDIA Jetson is that it simplifies the hardware requirements for implementing AI at the edge. Discrete GPUs and additional accelerators will continue to play a role in AI implementations, but Jetson provides a number of benefits for edge AI.

Purpose-built for edge AI

Discrete GPUs were originally designed to render complex graphics for immersive video games. They’ve since found new life powering a range of applications that require parallel processing. NVIDIA Jetson devices are specifically designed and optimized for AI edge computing, utilizing both CUDA and Tensor cores (depending on the model) to pack incredible analytical capabilities into a small form factor device.

Energy efficiency

NVIDIA Jetson modules are designed to be energy-efficient, making them suitable for deployment in power-constrained environments commonly found at the edge.

Traditional GPU and CPU combinations may not prioritize the same level of energy efficiency, as they are often used in data centers with more abundant power resources.

Compact form factor

Jetson modules are compact and intended for use in small form-factor systems, such as robots, drones, and IoT devices. This makes them well-suited for applications where space is limited.

Regular GPU and CPU combinations may not be as compact and may be designed for larger systems like desktop computers.

Software ecosystem

Jetson devices are supported by NVIDIA’s software ecosystem, including libraries and frameworks optimized for AI workloads. While regular GPU and CPU combinations can also run AI workloads, the software ecosystem may differ, and optimization for edge AI might not be as pronounced.

Benefits of NVIDIA Jetson

One of the most appealing elements of the Jetson platform is that it empowers users to deploy AI models, at scale, directly onto edge devices. That powerful combination of AI computation capabilities and on-site data gathering and analytics enables real-time insights without the need for constant connection to the cloud. Other benefits include:

  • Low latency: Because Jetson-powered edge devices process data on-site, they can help minimize the latency associated with sending data to centralized or cloud-based servers. Applications that can benefit from lower latency include manufacturing cobotics, autonomous warehousing equipment, and quality inspection, as well as anywhere decisions need to be made in the moment in order to optimize throughput and/or avoid injury.
  • Energy efficiency: A handful of power-hungry edge AI devices are one thing, but when scaling hardware deployments to hundreds or even thousands of systems, energy costs add up quickly. The Jetson platform is designed to deliver high performance computing capabilities while keeping power consumption to a minimum. Mobile deployments that may run on battery or solar power, or applications with otherwise limited access to power can also benefit from Jetson’s efficiency.
  • Versatility: When it comes to actually building solutions, Jetson supports a range of AI frameworks, and offers a robust developer library for users doing everything from basic image recognition to advanced natural language processing. Jetson is also impressively small, meaning Jetson-powered systems can be installed virtually wherever they’re needed.

Choosing the right Jetson device

NVIDIA offers a range of Jetson models, each designed for a particular subset of AI applications. The level of performance you require and the particular features you need will help guide your choice of Jetson platform. The most recent releases in the NVIDIA Jetson range were launched in 2023 as part of the Jetson Orin™ line.

Jetson Orin Nano™

This may be the “entry level” option within the Orin line, but it’s still quite a leap forward for Jetson. NVIDIA says the ultra-compact, 7-15W device provides up to 40 TOPS (Trillions of Operations per Second) of AI performance, which is up to 80X the performance of its predecessor, the Jetson Nano.

When to use Orin Nano:

  • Lightweight automation: Orin Nano is best suited for small-scale automation applications, particularly those with lower processing requirements. It’s a cost-effective option for users running simple workloads like basic quality control or monitoring.
  • Space constraints: Jetson Orin Nano’s small form factor makes it easy to integrate into space constrained applications or smaller edge solutions.
  • Prototyping: Orin Nano is also a good choice for prototyping or educational purposes. If you’re looking to experiment with AI in factory automation or test out the practicality of your implementation without a significant upfront investment, you may want to consider Orin Nano.

Jetson Orin NX™

The Orin NX steps up the performance to 100 TOPS of AI performance, but manages to maintain the same small form factor as the Orin Nano.

When to use Orin NX:

  • Everyday automation: Orin NX balances performance and power-use to provide a solid platform for mid-scale factory automation applications.
  • Vision-based applications: The processing capabilities of the Orin NX are well-suited to AI applications that rely on data from attached cameras. Tasks like object detection, measurement, tracking, and monitoring can all be deployed on an Orin NX system, as long as only moderate computing power is required.
  • Full factory deployment: For users looking to deploy a single hardware platform for a range of automation applications, the Orin NX is versatile enough to be used across an entire facility.

Jetson AGX Orin™

The powerhouse of the Jetson Orin line, the AGX Orin boasts 275 TOPS of AI performance in a slightly larger form factor that can still be integrated into the compact footprint of an embedded edge device.

When to use AGX Orin:

  • High-performance automation: AGX Orin is designed for compute-intensive factory automation scenarios. It’s capable of executing complex AI models and larger-scale data processing.
  • Robotics and advanced vision systems: If an automation deployment involves advanced robotics or cobotics, sophisticated computer vision, or complex AI algorithms, AGX Orin provides the necessary horsepower.
  • Large-scale edge AI deployments: AGX Orin is ideal for large-scale manufacturing facilities that require multiple, high-performance edge devices doing parallel processing and application coordination between a fleet of systems.

What does TOPS mean in AI?

The world of technology loves a unifying acronym to help compare hardware platforms, and AI hardware is no different. TOPS stands for Trillions of Operations per Second and has become one of the specifications most commonly used to illustrate a particular AI platform’s capabilities. TOPS indicates the number of computing operations an AI chip can handle. Of course, those “operations” can vary wildly, so like other overly simplified hardware specs (think GHz when measuring CPU performance) TOPS won’t always tell you how a particular system will perform when presented with a real-world workload. TOPS remains a relevant comparison metric, but it shouldn’t be relied on as the sole decision maker when choosing the right platform for your project.

Deploying Jetson in manufacturing

Manufacturers have been key in driving the demand for NVIDIA Jetson-powered devices. Forward-thinking businesses looking to optimize their production lines, increase throughput, and minimize scrap rates or downtime, are turning to the Jetson platform to bring AI to their facilities. Here are just a few examples of where Jetson has the potential to shine in a manufacturing setting:

Quality control and defect detection

When to leverage Jetson: Jetson devices are ideal for real-time quality control applications where quick decision-making is crucial. Edge computers powered by Jetson can process images and sensor data on-site, identifying defects and anomalies as products move down the production line.

Where to implement Jetson: Jetson devices can be installed at key inspection points on the line, such as assembly stations and packaging areas. The platform’s low latency can help identify defects quickly, helping to drive corrective actions and reducing the likelihood of faulty products reaching customers.

Predictive maintenance

When to leverage Jetson: Predictive maintenance can help manufacturers maximize throughput by preventing unplanned downtime. Jetson devices can analyze data from sensors to predict when equipment is likely to fail, enabling proactive maintenance.

Where to implement Jetson: Edge devices leveraging Jetson can be connected to, or installed on, machinery and equipment across the production floor. By continuously monitoring and analyzing tooling and equipment, manufacturers can schedule maintenance interventions before a failure occurs, minimizing disruptions.

Robotics and cobotics

When to leverage Jetson: Jetson provides the necessary computational power for AI-driven robotics for use in production automation. Tasks that require adaptability, precision, and the ability to learn from the environment can benefit from the power of on-site AI inferencing that Jetson offers.

Where to implement Jetson: Jetson systems can be integrated into robotic solutions handling tasks like pick-and-place, assembly, materials handling, and palletizing. The platform’s real-time processing capabilities enable robots to adapt to variations in the production environment and collaborate with human workers safely.

Process optimization

When to leverage Jetson: This is a broad bucket, but it’s worth highlighting that Jetson devices can optimize manufacturing processes across a facility or organization by analyzing data from various sensors and cameras. They can be particularly useful in complex production environments with multiple variables that can impact throughput.

Where to implement Jetson: Jetson devices can be installed at critical points in the production process to monitor and analyze data related to throughput, cycle times, and resource utilization. The insights gained can help manufacturers identify areas for improvement and enhance overall operational efficiency.

Here at OnLogic we offer a line of systems powered by NVIDIA Jetson SOMs designed specifically for industrial AI applications.

NVIDIA Jetson is making waves in the world of AI-enabled automation, providing a scalable and efficient solution for manufacturers looking to bring the power of AI to the edge. If you’re interested in seeing how Jetson-powered systems could benefit your business or have questions about which platform is right for you, reach out to our team or explore our line of systems built around the Jetson platform.

Darek Fanton

Darek is the Communications Manager at OnLogic. His passion for both journalism and technology has led him from the newsrooms of local papers to the manufacturing floor of IBM. His background in news gathering has him always on the lookout for the latest in emerging tech and the best ways to share that information with readers. In addition to his affinity for words, Darek is a music lover, juggler and huge fan of terrible jokes.

Published by
Darek Fanton

Recent Posts

Turning a Staircase into an Art Installation

At the heart of OnLogic’s headquarters is an incredible staircase - what our team of…

5 days ago

Edge AI Architecture: The Edge is the Place to be for AI

AI seems to be part of every technology-related conversation lately. While for some time AI…

1 month ago

Predictive Maintenance – how AI and the IoT are Changing Machine Maintenance

 The Internet of Things (IoT) and artificial intelligence (AI) are changing how we maintain equipment…

1 month ago

Partnering with UVM Students on Sustainable Innovation

Developing and acting on initiatives for sustainable technology solutions is a team effort. At OnLogic,…

3 months ago

What is OpenVINO and What are the Hardware Requirements?

Deep learning, a type of AI (artificial intelligence), is a subset of machine learning that…

5 months ago