UAV 2
UAV 1
UAV 3
MEC
Server
Base
Station
(a) MEC based Architecture [
73
]
(b) UAV Computational Offloading [
74
]
Core Network
Cloud
Storage & Data
Processing
Management &
Orchestrator
User3
User2
User1
4G/5G
WiMax
Wi-Fi
3G
(c) UAV-based IoT Platform
Figure 12:
Network Architectures of Cellular-connected UAV
according to mission’s policies and requirements. The pro-
posed implementation is shown in Fig.
11
. The deployment
occurs when the service requests are triggered to ESCAPE
as per requirement. OpenStack is used for running the cloud
environment and few laptop hosts are used as edge execu-
tion machines by Docker platform. High level commands
such as take-off, land, fly are used for controlling the UAV
behaviour from factory controller.
5.1.2. MEC Oriented Architectures
In general, UAVs possess physical constraints in terms
of computational capability, storage and battery capacity.
MEC has been identified as one of the promising techniques
to deal with the limitations of low computational capability
and restricted battery capacity of flying UAV. Some exam-
ples of resource-intensive tasks are trajectory optimization,
object recognition, AI processing in crowd-sensing. Due
to the limited onboard resources of the UAVs, computation
of above resource intensive tasks are not very efficient.
Hence, in such case, edge-cloud based network architec-
tures provide substantial improvements for operations of
cellular-connected UAVs.
In [
73
], the authors presented a UAV-enabled MEC ar-
chitecture applicable for cellular-connected UAVs. Fig.
12a
illustrates this architecture, where the UAV has some com-
putational task to be executed. This task can be offloaded
to the MEC server located with the ground station and, after
the computation, obtained results can be sent back to UAV
for their exploitation. Depending upon the volume of the of-
fload, there can be two modes of operation: (i) partial mode,
and (ii) binary mode. In partial offload mode, the whole task
is split into two parts. One part is executed locally and the
other part is executed by the MEC server (e.g., face recog-
nition use case). In binary offload mode, each task is exe-
cuted as one unit, irrespective of whether it is done locally
or at the MEC server (e.g., channel state information (CSI)
estimation). Both of these offload modes have advantages
and drawbacks. The selection of the suitable mode depends
on the nature of computational task being performed, UAV
structure and characteristics.
Considering the use case of trajectory optimization and
computational offloading in cellular-connected UAV, the
work in [
74
] presents a novel MEC setup, where the UAV
needs to offload some of its processing task to the ground
station. The UAV flies from an initial location to a desti-
nation location and offload the task to selected ground base
stations during the trajectory. The goal of the MEC setup
is to minimize the total time for UAV mission considering
the maximum speed and ground station capacity constraints.
This setup is shown in Fig.
12b
.
In reference work [
75
], the authors proposed a 5G net-
work slicing concept extend to video monitoring with UAVs
having MEC facilities. The surveillance area is divided into
multiple zones and a set of UAVs are assigned the task to
monitor a specific zone. The MEC enabled UAVs could of-
fload the captured data and video streams with acceptable
quality and performance.
Do'stlaringiz bilan baham: |