Page 38 - EE Times Europe Magazine | April2019

38 EE|Times EUROPE — Boards & Solutions Insert

SPECIAL REPORT: AI AT THE EDGE
A Much Smarter Edge for IoT

By Nitin Dahad

he internet of things has been touted
for several years as the answer to
many challenges. Connected IoT
Tdevices can improve efficiencies and
productivity in industrial systems, provide
valuable feedback mechanisms for connected
health-care systems and wearables, and
provide a vast array of capabilities to improve
driver assistance and enable a path toward
autonomous vehicles.
Fulfilling the promise of IoT has relied to a
large extent on sensors in a network collecting
data that is transmitted via a gateway to the
cloud, processed, analyzed, and returned to the
local system or sensor as actionable feedback.
Over the past couple of years, however,
developers and systems integrators have come
to realize that this approach poses issues
around latency, data security, and bandwidth cost. Adding more intel- the edge of the gateway? These are some of the questions we asked
ligence at the edge can mitigate those challenges by speeding response for this special report as we looked for points of consensus among the
times, keeping data secure and private, and minimizing data commu- vendors supplying the chips and systems.
nication costs. So putting intelligence at the edge is a no-brainer, right? We also explore how much intelligence should be added to the edge
It is — but there are, of course, practical limitations. and what the practicalities and limitations are for doing so. And we
As edge intelligence has become the buzz phrase of the embed- look at some of the top edge-AI chip startups and make some predic-
ded-systems industry, its definition has gotten rather fuzzy and now tions for the key trends coalescing around AI inferencing. ■
spans a wider swath of the network. Some define the edge as anything
not in the cloud; but even within the edge, are you at the endpoint or Nitin Dahad is a staff correspondent at AspenCore.

SPECIAL REPORT: AI AT THE EDGE
Let’s Talk Edge Intelligence

By Nitin Dahad

hat’s the difference between edge and endpoint? It and others are using the term for a small energy-harvesting sensor
depends on your perspective. with low-power wireless connectivity. Edge is used in relative ways and
When new industry buzzwords come up, the challenge distinguishes the more local from the more central.
Wfor people like us who write about the topic is figuring “But indeed, there is a need to distinguish
out what exactly a company means, especially when it uses the term between the various kinds of things that you
to fit its own marketing objective. The latest buzzword is actually a find at the edge. Sometimes, you hear terms
phrase: edge AI. like ‘edge of the edge’ or ‘leaf nodes.’ Edge
Because of the proliferation of the internet of things and the ability AI can be many things, including a compute
to add a fair amount of compute power or processing to enable intelli- server in a car.”
gence within IoT devices, the “edge” can be quite wide and could The key, he said, is that “endpoint AI
mean anything from the edge of a gateway to an endpoint. resides at the location where the virtual
We set out in search of industry consensus on the definitions of world of the network hits the real world,
edge and endpoint, who might want to add artificial intelligence at Wolfgang Furtner, where sensors and actuators are close.”
the edge, and how much “smartness” could be added. Infineon It’s all about semantics and where you
We discovered that the answers depend on your perspective. But, draw the boundary, according to Markus
for starters, essentially anything not in the cloud can be defined as Levy, director of machine-learning technologies at NXP Semiconduc-
the edge. tors. “Edge machine learning [ML] is the same as an ‘endpoint’ machine
learning, except edge ML can also include ML that takes place in a
MANY ‘EDGES,’ BUT JUST ONE ENDPOINT gateway or even fog compute environment,” said Levy. “Endpoint ML is
Probably the clearest definitions of edge and endpoint came from Wolf- typically related to distributed systems — for example, where our cus-
gang Furtner, Infineon Technologies’ senior principal for concept and tomers are adding intelligence even down to the sensor level. Another
system engineering. “The term ‘edge AI’ inherits its vagueness from example is a home automation system, where there are ‘satellite’
the term ‘edge’ itself,” he said. “Some people call a car an edge device, devices, such as thermostat, doorbell camera, security cameras, or other

APRIL 2020 | www.eetimes.eu