Der Blätterkatalog benötigt Javascript.
Bitte aktivieren Sie Javascript in Ihren Browser-Einstellungen.
The Blätterkatalog requires Javascript.
Please activate Javascript in your browser settings.
International 2020 Elektronik 15 Microcontroller The tech sector thrives on acronyms Heres one thats fastgaining traction The AIoT It represents an Internet of Things IoT enhanced by Artificial Intelligence AI This isnt technology marketers inventing a new term for the sake of it it is the description of a new compute evolution where IoT devices are being rapidly supplemented with AI capabilities AI is a force multiplier for embedded and IoT deployments driving advanced processing down into endpoint devices and enabling them to do far more Gartner predicts that by 2022 more than 80% of IoT projects in the enterprise will include some form of AI today this is figure is nearer 10% There are many examples of endpoints employing AI already notably smartphones running ondevice capabilities such as facial recognition for handset unlock or categorizing photos Voice recognition is also rapidly becoming a classic AIdriven feature both for smart phones and smart speakers For IoT to scale and for companies to take advantage of the enhanced insights and experiences offered by advanced AI small IoT devices must become more capable of ondevice processing shifting compute closer to the source of data The first generation of smart speakers typically used a highperformance application processor such as an Arm Cortex-A CPU to handle the wake word then offload the speech processing to a data center Smart speaker technology has progressed to using a microcontroller processor such as an Arm Cortex-M CPU for the wake word coupled with an applications processor for sending the speech to the cloud to process Amazon for example recently announced their Alexa Voice Service AVS integration for AWS IoT Core making it easier and more costeffective for developers to add Alexa Builtin capabilities to small devices powered by Arm Cortex-Mprocessors however the bulk of the processing is still done in the cloud Smart speakers are almost entirely supplied by one of the Big 5 tech companies so they have the data center resources needed to support what could amount to many millions of requests per second from a global customer base Some companies are cutting the cord to the cloud even further such as Snips an Arm partner and member of its Innovator Program Snips has developed voice recognition technology that runs at the endpoint recognizing wakewords and a host of other commands locally on the device using both microcontroller and application processors Snips says there are many benefits from taking this AIoT processing approach notably security privacy and faster response times since there is less reliance on the cloud and gateway Also because an AIoT system doesnt rely on external connectivity it is more resilient easier and cheaper to scale In the future technologies like Arm Helium that bring an uplift in digital signal processing DSP and machine learning ML for future Cortex-Mprocessors will enable endpoint AI devices such as smart speakers with even more ondevice processing capability and make it even simpler for software developers to implement Distribution of data processing from cloud to endpoint While there are many many companies deploying IoT there is a strong underlying cohesiveness driven mainly by the Arm architecture upon which most devices tend to be based The AIoT will have the same characteristics so as the ease of adding intelligence to IoT devices increases we can expect competition in end products to rise in nearly every market The nature of how the AIoT will work in practice is also in line with the muchneeded distribution of computing We already see more compute power moving from the datacentre into the network at the furthest edge where network gateways sit Beyond that we also see AI moving out into endpoint devices to improve decisionmaking latency and take cost out of the system This new model will not replace the old one entirely but will complement it by spreading computing process between cloud edge and endpoint This hybrid system will be far more sustainable in resilience terms and in literal sustainability terms by reducing the energy footprint of our industry Figure 1 Endpoints such as smart phones are using AI to change the way we interact with technology photo Arm Figure 2 Aquaseca shows the impact an endpoint device using ML can be to help realworld issues photo Arm