When we talk about “Edge Computing,” we are often talking about PCs, but PCs are a tiny part of the edge that will enable the coming AI world. The real volume will be in sensors that give the deployed AIs the ability to sense and interact with the world around us. These sensors increasingly need to be intelligent so that the amount of data aggregation in Cloud AI data centers can be reduced to manageable levels and so that latency doesn’t render large-scale AI deployments and upgrades ineffective.
At embedded world this month, Qualcomm made a huge step by announcing breakthrough Wi-Fi technology and introducing new AI-ready IoT home and industrial platforms.
Let’s talk about that this week.
The Critical Need for Smart Sensors (including cameras)
One of the major components of the coming AI wave are digital twins that will be used to create digital clones of real things in the metaverse. These clones can then be used in simulations to assess what is possible, manage remote resources, and to forensically analyze a past problem or event. The problem that needs to be addressed is that to be effective, the digital twin must remain synchronized with its physical counterpart, or the simulation, management action, or past event analysis will become unreliable and increasingly no longer a twin at all.
The larger scale the project, the greater the need for dependable, connected sensors that can provide telemetry real time and inform the centralized AI so that it can make both timely and accurate decisions.
This means that the devices need to be similar (standards based) to ensure both interoperability and rapid replacement in case of a problem, and they need to be performant both in terms of networking capability (so they can report) and processing capabilities so that the data stream from them is manageable in aggregate without bottlenecking the data link.
This combination of capabilities, if done right, should reduce the cost of deployments while increasing their accuracy and reliability. We’ve just been waiting for some company to step up.
Qualcomm Steps Up to IoT
That is exactly what Qualcomm did when they announced new industrial and embedded AI platforms along with a powerful Wi-Fi SoC that is incredibly power efficient and reasonably priced, resulting in an initial 35 companies that will build solutions based on this technology. Qualcomm also announced a unique QCS6490 processor (Qualcomm® Robotics RB3 Gen 3 Platform) optimized for premium tier IoT devices with support for Wi-Fi 6e. These devices will have the performance and battery life needed for deployment at scale to assure the connected AIs have the timely information they need to fulfill their mandate.
Coupled with Qualcomm’s 670 CPU and a Qualcomm Hexagon processor fused AI-acceleration, this announcement not only showcases that Qualcomm knows what is needed for AI IoT but has stepped up to effectively provide it.
Potential solutions using this technology are impressively broad and range from security cameras and rugged handhelds and tablets to purpose-built industrial and commercial IoT applications including scanners, dash cameras, handheld computers (including those that are ruggedized), tablets, kiosks and point-of-purchase hardware.
This platform will support up to a whopping 5 concurrent cameras, Wi-Fi 6E and accelerated AI and, according to Qualcomm, it will have long lifecycle support for multiple OSs, hardware, and software. Finally, the platform will support up to 3 cameras per module, significantly reducing the cost of any AI solution that requires multiple camera monitoring.
Wrapping Up:
The rush to AI has opened up a number of opportunities. One is the creation of sensors that not only can convey what is going on to remote AIs, but also keep the increasing number of digital twins synchronized with their real-world counterparts to better assure the resulting virtual constructs (like Earth 2).
The success of our AI future will depend on efforts like this to ensure our AI present is based on accurate IoT sensors based on technology like this.
