Unlocking the Potential of Edge Computing

edge computing

Edge computing offers several key advantages that cloud computing can’t match

Edge Computing has the potential to be revolutionary. Moving processing and intelligence away from the cloud, and close to IoT devices, can deliver huge benefits in the way data can be acted upon, changing the way that business is done.

The main benefits are twofold: increased speed and a reduced requirement for a fast network connection. With the first, data no longer needs to be sent to the cloud for processing, so results are almost instantaneous. But it’s the second benefit that’s perhaps more interesting. Rather than piping data to the cloud, edge devices can make decisions offline. For areas of poor or unreliable connectivity, ‘Edge AI’ systems can work where cloud systems would fail.

A prime example of this is autonomous cars, where decisions need to be made instantly and often in areas without a solid network connection. Here, Edge Computing (coupled with Edge AI) is the only way for situational decisions to be made. In fact, it’s been calculated that Autonomous Emergency Braking (AEB) systems, which can apply a car’s brakes when an imminent collision is predicted, have reduced crashes by 38%.

What do you need for edge AI?

Edge AI implementation requires IoT sensors that can transmit data back to a localised computer. The computer then uses the intelligence programmed into it to make sense of the input so it can be acted upon.

Of course, using computers close to the point of data collection requires robust, powerful and slimline embedded platforms, such as the VIA SOM-9X20 module and the VIA ALTA DS 3 system.

edge computing, Autonomous cars on the road
Distributed Edge Computing will allow autonomous cars to respond faster to the demands of unpredictable roads.

As simple as that sounds, the implementation of Edge AI can range from the simple to the exceptionally complex. The truth is that there’s no one-size-fits-all solution for Edge AI and this has led to several common barriers to implementation, which are neatly summarised by a Penton survey. Here, we’ll look at three of the biggest barriers and how they can be overcome in different industries.

1. Data privacy and security

Privacy is high-up on everyone’s list, along with the protection of personal data. Consumers are becoming more aware of privacy issues and interested in how their data will be used, while tougher data protection laws, such as the EU’s GDPR, are making companies think twice about how they collect, store and use information.

Yet, Edge AI can help improve privacy and security. Edge computing systems can be trained to filter out private information, formatting data collection for big data analysis to make it generic. And, without a single point of storage for the pre-formatted data, security is subsequently boosted.

A facial recognition system used to as part of a customer engagement platform in retail, for example, can identify a shopper using local processing resources, so no private information is sent to the cloud. And, for data collection, only generic demographic information, such as age and gender are collected.

For healthcare, data can be stored and processed locally too. So, an ECG monitoring machine could only send an alert if an unusual rhythm is detected, rather than storing all data in the cloud. Likewise, autonomous cars will react to stimulus, but may only send generic driving stats back to the cloud to update maps or refine existing automated systems.

2. High cost of implementation and legacy systems

Budgets are always tight in business, and there’s a genuine concern about implementing new technology and how long it will take to pay back the investment. While a full cost analysis should always be performed, price shouldn’t be a barrier to entry. Off-the-shelf systems and smaller trials can help justify any spend.

In retail, VIA’s Smart Retail Engagement System gives you the ability to identify VIP customers, blacklist known shoplifters and collect customer demographics.

edge computing, Woman in a store scanned by facial recognition
Combining Edge Computing technology with facial recognition can arm retailers with valuable data-driven insights.

Likewise, the VIA Mobile360 ADAS system can be fitted to existing vehicles, enhancing safety with 360-degree camera views and collision warnings.

In the manufacturing sector, could you add additional sensors to legacy equipment and then use Edge AI to monitor for potential problems? In that way, you can work with existing equipment but benefit from predictive maintenance schedules, reducing downtime. Intel managed to increase uptime in its fan filtration units by a whopping 97% using vibration sensors to identify faults.

3. Inadequate infrastructure

IoT devices can generate a huge amount of data, all of which needs to be handled, managed and controlled. There’s a concern about the level of infrastructure that this requires, with many businesses already finding their networks operating at full capacity.

Yet, Edge Computing can address this. By monitoring and controlling data locally, Edge AI devices can restrict the amount of information that’s collected and transmitted. As a result, only data that’s genuinely useful for cloud-based big data analytics projects needs to be collected and stored centrally. And, the scheduling of data transmission can be handled carefully, ironing out any problems with slow or intermittent network connections.

Edge AI has the power to transform businesses, yet the process needn’t be complicated or expensive. By focusing on a single project or area, Edge AI can be introduced at a pace and cost that matches your business needs now and into the future.

VIA has developed a growing range of highly-customizable Edge AI solutions to help you accelerate the adoption process, including the VIA SOM-9X20 module, the VIA Edge AI Developer Kit, and the VIA ALTA DS 3 system.

VIA Technologies, Inc.