Page 77 - ITU Journal Future and evolving technologies – Volume 2 (2021), Issue 2

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 2

Wu. et al. found that video analytics done in the edge
instead of on the cloud reduces the latency up to 61%
using 4G [17], expressing the usefulness of edge

computing in these situations. The analysis can be done
entirely in the edge servers, but as they are capacity

limited it may be done only partially, the rest being
handled in the cloud. Some previous work leverages the
edge computing power to ilter images taken from a
disaster context and send only relevant ones to the

cloud for advanced processing. Indeed, several images

and video are taken by smartphones [16], drones [34]

or cameras [17], and by sending them all the cloud will

burden the network, as they do not all contain useful

information. The iltering will save precious bandwidth

and act in real time without human intervention. COCO,
Fig. 1 – The multiacess edge computing paradigm
proposed by Zhao et al. [35], is a MEC‑based adaptive
types of services to carry out their crucial objectives that image sensor that uploads to the cloud only images with
involve life and properties. Thus, in the next section, we speci ic content. Liu et al. [16] propose Echo that is an
are going to review two main use cases, disaster manage‑ edge‑based face recognition frame‑ work, that also
ment and military and what kind of edge services they ilters images and preprocesses them before sending

may request. We will also review the technical speci ics them to the cloud. Chemodanov et al. [30] propose

of these two uses cases. geospatial video analytics that analyze images from

many devices in a broad area to deliver information to

rescuers. They employ fog computing to preprocess

3.1 Disaster management

images and manage human‑computer interactions, that
With climate change, disaster occurences are bound to are trivial tasks with low latency expectations. edge
increase [31], having important social and economic im‑ computing allow other applications, like localization and
pacts. The rescuers need to be prepared and supported path inding for autonomous agents (drones and boats)
ef iciently to carry out their mission, saving many lives when searching for victims in the sea [36]. Avgeris et

and recover from the situation. IoT is recognized as a al. [37] propose SMOKE, a three‑layered cyber‑physical

relevant technology for providing useful support to res‑ social system to detect forest ires and assist public au‑
cue operations [2, 28]. But all the data produced by IoT thorities. They use the edge layer to process images cap‑

needs low‑latency processing to be useful for rescuers in tured by IoT nodes to detect ire at an early stage. They

real time. For this, MEC is a promising candidate due to also propose the horizontal and vertical scaling of the

its proximity [5] and its on‑site rapidly deployable [32, edge resource to adjust the QoS. It can monitor rescuers’
33] networking structure, even in dif icult environments. health in mission and alert about their state [17]. Finally,
In the next section, we review disaster management ap‑ arti icial intelligence which runs at the edge provides an
plications that use MEC but also other edge computing action plan and helps decision‑making [17].
paradigms as currently there is little work only on mobile
MEC. 3.1.2 Technical speci icity
3.1.1 Edge‑enabled disaster management ap‑ Architecture Several architectures are “three‑layered”

plications where the irst layer is composed of ield sensors, then
the second edge layer and the third one that is the cloud.

Edge services enable low‑latency applications that pro‑ The cloud is kept to store historic data [17] and runs less

vide situation awareness to rescuers. These types of ap‑ time‑sensitive or heavy tasks [35, 16, 30]. Edge servers
plications deliver important information about the situa‑ are often in mobile units near the disaster scene like ire

tion, helping rescuers adapt and organize their mission. vehicles, public buses [38] or drones [36]. It highlights

A common use of edge computing by disaster manage‑ the need for edge resources mobility management to sup‑

ment applications is for video and image analytics [17, port the rescuers as they move on the ield. Sensor de‑

16, 34, 17, 35, 30, 36, 37] as edge computing responds vices that gather data are heterogeneous. Wireless wea-

to the short response time requirements of these appli‑ rable sensors monitor health or help for localization. Body

cations. It helps discover victims’ locations, count and cameras [17] and smartphones [35, 16, 30], from civilians or

state and provide facial recognition for missing persons. rescuers, capture the environment to analyze paths,

It also helps analyzing the environment to detect danger‑ recognize missing people or help evaluate their state and

ous paths clogged with ire or hazardous chemicals and injuries. Surveillance cameras [30, 17] are also used to
paths obstructed by wreckage. evaluate the environment on a larger scale.

© International Telecommunication Union, 2021 63