NXP Semiconductors has recently teamed up with Amazon and extended Amazon Web Services’ (AWS) cloud capabilities to local devices, to enable edge processing, which NXP believes will be the next level of artificial intelligence in the smart home setting.
NXP Semiconductors has recently teamed up with Amazon and extended Amazon Web Services’ (AWS) cloud capabilities to local devices, to enable edge processing, which NXP believes will be the next level of artificial intelligence in the smart home setting.
Edge computing is a new class of gateways or systems that performs data processing near the source of the data, or at the edge of the network, working in conjunction with the cloud.
For Amazon’s IoT services, its cloud software can also be extended to run on local devices. One application is that a camera collects data which is sent to an IoT gateway for edge processing, before it is sent to the cloud for further analytics.
“Because it is impractical to connect all those IoT devices directly with the cloud,” said Sam Fuller, Head of Strategy System Solutions of Digital Networking at NXP Semiconductors. There are connectivity cost, latency issue, etc. “Most of the stuff you do, you don’t need to go to the cloud,” he said. This applies to a smart home setting too. Amazon Echo does some local processing. “I don’t want to just send everything, maybe for privacy or bandwidth reasons,” Fuller explained.
Sam Fuller, head of strategy
system solutions, digital
networking at
NXP semiconductors
“When you start to upload video to the cloud, you quickly find out it’s a lot of data. IoT devices can generate lots of data, will require a lot of intelligence. Edge computing is designed to solve the problem,” Fuller noted.
The edge device – which is often times an IoT gateway – is fairly low-power, light-weight, but runs the same software as the cloud, Fuller pointed out, adding that NXP makes gateway processers that have the capacity.
It is also about aggregating IoT data from many users or sensors. In a wind farm, for example, a lot of data – such as wind speed, turbine conditions and amount of power generated – is created. There can be thousands of windmills, and it would be utterly impractical to connect all the data directly to the cloud. Local systems are therefore set up to process the data.“
Another example is surveillance cameras that record videos locally. The cameras don’t stream directly to the cloud. Typically the cameras will have some edge intelligence, which, for example, may allow them to recognize the outline or shape of a gun carried by someone in a public place. Upon detection, this finding will be sent to the cloud for further processing.
Fuller projects more adoption of edge computing within homes in the near future. Many companies have come up with interesting ideas on how to leverage edge computing, according to Fuller, who added, “it’s about moving to the next level of artificial intelligence in homes.”
Smart home technology continues to improve
NXP pays much attention to smart home because its sell semiconductor solutions that go to smart devices like sensor, thermostat and remote control. The homes will get smarter and smarter. We are not even close to the end of what ultimately can be accomplished by smart homes, Fuller said.
Two main obstacles lie ahead of broader adoption. The first is the economics, or the cost of going to smart home. However, the cost is bound to decrease as processing performance advances.
The evolution of computers is a case in point. A mainframe computer used to cost US$20 million, but right now a small access point that has more processing power and bandwidth costs less than $20. “From $20 million to $20, that has been the trend of electronics, we are just starting to see what can be done in the smart home, as we continue to drive the cost down on technology,” Fuller said.