Implementing NVIDIA Jetson for AI-powered automation

NVIDIA® Jetson™ has emerged early on as a leader in the ongoing race for hardware platform superiority to support the tremendous growth of artificial intelligence (AI) deployments. NVIDIA is already a household name in the world of advanced GPU-technology. The suite 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, we take a closer look at the value that NVIDIA Jetson offers for those implementing AI solutions. We talk about the key differences between versions of Jetson and outline when and where Jetson is best deployed for industrial automation applications.

Jetson versus discrete GPU and CPU

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

Built specifically for AI at the edge

Discrete GPUs were originally designed to display complex graphics for immersive video games. Since then, they have found new life in powering a range of applications that require parallel processing. NVIDIA Jetson devices are specifically designed and optimized for AI edge computing, using both CUDA and Tensor cores (depending on the model) to integrate incredible analytical capabilities into a compact device.

Energy Efficiency

NVIDIA Jetson modules are designed to be energy efficient. That makes them suitable for deployment in power-constrained environments. Something common at the edge of the network.

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

Compact size

Jetson modules are compact and intended for use in small devices such as robots, drones and IoT devices. This makes them well suited for space-constrained applications.

Ordinary GPU and CPU combinations are not always so compact and are often designed for larger systems such as desktop computers.

Software ecosystem

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

Benefits of NVIDIA Jetson

One of the most attractive elements of the Jetson platform is that it allows users to create AI models at scale directly on edge devices implementation. This powerful combination of AI computing power and on-premises data collection and analysis enables real-time insights without the need for a constant connection to the cloud. Other benefits include:

• Low latency: Because Jetson-powered edge devices process data onsite, 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 warehouse equipment, and quality inspection. In fact, anywhere decisions need to be made to optimize throughput and/or prevent injury.
• Energy Efficiency: A handful of power-hungry AI devices is one thing, but when hardware deployments are scaled to hundreds or even thousands of systems, energy costs add up quickly. The Jetson platform is designed to provide high-performance computing capabilities while minimizing energy consumption. Mobile deployments that can 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. It also provides a robust developer library for users in areas ranging from simple image recognition to advanced natural language processing. Jetson is also impressively small. That means Jetson systems can be installed virtually anywhere they are needed.

Choosing the right Jetson device

NVIDIA offers a range of Jetson models, each designed for a particular subset of AI applications. If you know what performance level and specific features you need, choosing the right Jetson platform is a lot easier. The most recent releases in the NVIDIA Jetson series were introduced in 2023 as part of the Jetson Orin™ line.

Jetson Orin Nano™

This might be considered the entry level of the Orin line, but it is still quite a leap for Jetson. NVIDIA says says the ultra-compact 7-15W device delivers up to 40 tops (trillions of operations per second) of AI performance. This is 80 times the performance of its predecessor, the Jetson Nano.

When to choose the Orin Nano:

• Lightweight Automation: Orin Nano is best suited for small-scale automation applications, especially those with lower processing requirements. It is a cost-effective option for users performing simple workloads, such as simple quality control or monitoring.
• Space limitations: Thanks to the small size of the Jetson Orin Nano, you can easily integrate this system in applications with limited space or in smaller edge solutions.
• Prototyping: Orin Nano is also a good choice for prototyping or educational purposes. Want to start experimenting with AI in factory automation or test the workability of an implementation without a significant upfront investment? Then consider the Orin Nano.

Jetson Orin NX™

The Orin NX increases performance to 100 tops AI performance while maintaining the same small size as the Orin Nano.

When to choose the Orin NX:

• Daily automation: Orin NX gives a good balance between performance and power consumption to provide a solid platform for medium-sized factory automation applications.
• Vision-based applications: The processing capabilities of the Orin NX are well suited for AI applications that rely on data from connected cameras. Tasks such as object detection, measurement, tracking and monitoring can all be implemented on an Orin NX system as long as no more than average computing power is required.
• Full factory implementation: For users who want to implement a single hardware platform for a range of automation applications, the Orin NX is versatile enough to be used throughout a facility.

Jetson AGX Orin™

The AGX Orin is the powerhouse of the Jetson Orin series, delivering 275 tops of AI performance in a slightly larger size, but which can still be integrated into the compact dimensions of an embedded edge device.

When to choose the AGX Orin:

• High-performance automation: AGX Orin is designed for computationally intensive factory automation scenarios. It allows you to run complex AI models and large-scale data processing.
• Robotics and advanced vision systems: If you need advanced robotics or cobotics, advanced computer vision or complex AI algorithms for an automation implementation, AGX Orin provides the power you need.
• Large-scale AI implementations at the edge: AGX Orin is ideal for large-scale manufacturing facilities that require multiple, high-performance edge devices for parallel processing and coordination of applications across a range of systems.

What does tops mean in AI?

In the world of technology, they do like acronyms to compare hardware platforms, and this is no different for AI hardware. Tops stands for Trillions of Operations per Second and has become one of the most widely used specifications to illustrate the capabilities of a particular AI platform. Tops indicates the number of computing operations an AI chip can handle. Of course, these "operations" can vary tremendously, so like other, slightly oversimplified hardware specifications (think GHz when measuring CPU performance) tops don't always tell you how a given system performs when faced with real-world workloads. Tops remains a relevant comparison metric, but you should not use it as the sole criterion when choosing the right platform for your project.

Implementing Jetson in a production environment

Manufacturers have played a key role in driving demand for NVIDIA Jetson devices. Forward-thinking companies looking to optimize their production lines, increase throughput and minimize waste or downtime are choosing the Jetson platform to introduce AI into their facilities. Here are some examples of where Jetson can excel in a manufacturing environment:

Quality control and error detection

When to deploy Jetson: Jetson devices are ideal for real-time quality control applications where fast decision-making is critical. Jetson-powered edge computers can process images and sensor data on site. Thus, defects and deviations can be identified as products pass through the production line.

Where to implement Jetson: You can install Jetson devices at key inspection points along the line, such as assembly stations and packaging areas. Thanks to the platform's low latency, you can quickly identify defects and take corrective action. This reduces the likelihood of defective products reaching the customer.

Predictive maintenance

When to use Jetson: With predictive maintenance, manufacturers can maximize throughput by avoiding 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 using Jetson can be connected or installed on machines and equipment throughout the production floor. By continuously monitoring and analyzing tools and equipment, manufacturers can schedule maintenance interventions before a failure occurs. This minimizes interruptions.

Robotics and cobotics

When to deploy Jetson: Jetson provides the computing power needed 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 inference that Jetson provides.

Where to implement Jetson: Jetson systems can be integrated into robotic solutions to handle tasks such as pick-and-place, assembly, material handling and palletizing. The platform's real-time processing capabilities allow robots to adapt to variations in the production environment and work safely with employees.

Process optimization

When to deploy Jetson: This is a broad area, but it is worth emphasizing that Jetson devices can optimize production processes in a facility or organization by analyzing data from various sensors and cameras. They can be particularly useful in complex manufacturing environments with multiple variables that can affect throughput.

Where to implement Jetson: Jetson devices can be installed at critical points in the manufacturing 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 improve overall operational efficiency.

OnLogic systems with NVIDIA Jetson SOMs are designed specifically for industrial AI applications.

NVIDIA Jetson is creating a furor 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. Want to know how Jetson-powered systems can also be deployed in your business, or have questions about which platform is best suited? Then contact our team or explore explore our line built around the Jetson platform.