Edge AI: What is it and how Does it Work?

Edge AI, what is it? How does it work? What are the benefits? It was the subject during one of our OnLogic Live sessions and in this blogpost we’ll take a deeper dive. We’ll explore what AI (artificial intelligence) at the edge means, review some edge AI use cases, and cover some concepts that hopefully help you gain a better understanding of the benefits of edge AI. 

What is edge AI?

Edge AI is the use of artificial intelligence at the edge of a network. Since it is happening near where the data is created, it can provide real-time actionable insights. AI calculation can, of course, take place in a centralized cloud or offsite data centers. However, putting AI capabilities at the network’s edge helps to speed decisions and actions based on the information gathered from on-site sensors.

Real examples of edge AI

A great example for edge AI are autonomous vehicles. Check out our customer story on Bear Flag Robotics about their autonomous tractors. This solution uses information from a myriad of inputs including: 3D cameras, LiDAR and position sensors to know where to turn to avoid uneven soil, how to adjust equipment, and when to brake to avoid obstacles. 

These are all decisions that need to happen quickly – there isn’t time to send the data to the cloud, process it, and then return instructions to the tractor. Having the data ingested and processed right on the computer mounted on the tractor rather than in the cloud, eliminates that delay, called latency. 

Not to mention, access to an internet connection might be a challenge on a farm field. With AI at the edge, the computing can just continue, regardless of where that data lives. 

Benefits of using artificial intelligence at the Edge

You might recall our blog post about the benefits of Edge Computing which reviews the benefits for an industrial automation use case. The benefits are similar for an AI use case including: 

• Lower latency
• Increased speed
• Ability to get more data at faster rates
• Reduced bandwidth utilization
• Increased data processing, analysis, and analytics

Since decisions can happen in real-time based on real-time data, AI enables a speedy response. 

Hardware requirements for edge AI

So, now that you understand the benefits of edge AI, you might be wondering – what hardware do I need to truly make it successful? The truth is, it depends. Different applications will have different hardware requirements. There are a wide range of considerations. 

Industrial I/O

Your edge computer may need to connect to many IoT devices. Take inventory of your connected devices to ensure you are configuring your solution with the appropriate industrial I/O with room for expansion. For example, the OnLogic HX500 has 8 USB ports, dual LAN, and an optional COM port for legacy connections.

Location of your edge AI application – environmental considerations

One of the most important considerations is the location of your application. There are a lot of external forces that can have an impact on your computer. You might have an application that is constantly prone to dust, dirt and other particles that could harm a traditional computer. In that case, an industrial fanless computer is a must. 

Think about the environment experienced by Bear Flag’s autonomous tractor. Not only is a tractor-mounted computer exposed to dirt, but it is constantly affected by impact forces and vibrations. A traditional computer probably couldn’t handle that environment and might experience a loss of data. That would prevent your AI from making the correct decision at the right time. In this situation, a rugged computer might be your best bet. 

How to handle the amount of data

Another consideration is the amount of storage on your computer. Data is the main source of nutrition for an AI. All the sensors and data streams that your edge computer is connected to can pile up data fast. What about situations where your edge computer is disconnected from a network and won’t be able to communicate data to a fog or cloud layer? This could be the case for a mobile AI application. There are two ways to go: storage and wireless connectivity. 

Storage

You could add a lot of physical storage on your computer that acts as a buffer or cache, that syncs to the cloud or your data units as soon as you are connected to a network again. Storage solutions based on NVMe and PCIE 4.0 ensure that you can still pack a more than decent amount of storage in your industrial/rugged computer while still benefiting from its compact size. 

Connectivity capabilities

You could add wireless connectivity capabilities to your computer solution, so that you can always maintain a connection to your cloud to store and retrieve data. A computer with 4G connectivity relies on a SIM card that will give you data connectivity in order to communicate with different platforms or servers. However, you can also decide to take both options: ensure connectivity on your computer in any situation, but also configure enough storage to act as a buffer in situations where you don’t have connectivity. 

Visual processing

What if your AI Application requires visual processing? For example: you have an application that has a camera to scan its surroundings and processes visual input based on images or videos. In order to make sure your AI application runs as smoothly as possible, you need to consider the graphics processing capabilities of your edge AI computer. In a lot of cases, the newest processors have the computational power to handle the graphics processing work that needs to be done for your application such as those found with our Karbon 800 Series. However, in some cases, your computer might require some heavier-duty visual processing, which could require an industrial computer with a dedicated graphics processing unit. 

Are you ready for AI at the edge?

Whether you are an engineer looking at the best way to implement AI, or a human superhero flying through the galaxy with a billion-dollar space ship that benefits from a super advanced AI on-board to make all important decisions, there’s no denying that AI at the edge is becoming essential in today’s technological world. As technology is becoming more flexible, and as applications require more and more real-time data, AI at the edge is going to become one of the most essential concepts to make artificial intelligence succeed. 

Are you ready to start your edge AI journey? We’re more than happy to advise you on the hardware requirements for your AI application. Download our checklist for choosing an industrial computer or reach out to one of our solution specialists to choose the perfect computer for your next AI project.