IEEE Consumer Electronics Magazine - January/February 2023 - 46

An SDN-Based Framework in UAV Networks
To address these challenges, centralized control,
external computation, and programmability
are required, which is provided by SDN. As SDN
is a centralized architecture, thus, all the computation
is shifted to the controller, which
reduces the power consumption of UAVs in SDNbased
UAVNets. In addition, radios can be
turned off for power conservation. Furthermore,
the global view of the network makes path selection
simple and straightforward.15 Since SDN
provides an elastic and programmable network;
therefore, programmability helps to reduce collision
risks. However, applying SDN in UAVNets
introduces new challenges, such as uneven load
over the links, energy consumption, and collision
among UAVs.
Problem Statement
In SDN, only programmable switches are used
to shift decision-making to a centralized controller.
These switches provide their information to
the controller. In response to this information, the
controller generates a global view of the network
and installs flow rules on underlying devices.
Once a flow rule is installed, devices forward data
according to these rules till the end of the flow. In
the case of multiple paths, there can be an uneven
distribution of load. This situation also results in
UAVs' battery exhaustion, which ultimately needs
to be replaced.
Assume that a network has a topology graph
G that can be represented as G=(V;E), where V
= fv1;v2;v3...vmg represents UAVs, and E = fe ¼
ðu; vÞ : u; v 2 Vg is a set of edges to connect V
number of UAVs. Any path between consumers
(i.e., a source s and destination t) is represented
as Pt
s = fvs;v1;v2...vtg. Notably, traffic originator
is a consumer, however, from a controller's perspective,
it is UAV from where the flow begins.
Hence, UAVs are used to represent source and
destination. Moreover, every node V has a
weight We, where e 2 [1-M] is a series of M nonnegative
weights or cost functions. Finally, the
total weight of path pi can be calculated as;
Wpi
=
Pm
en¼0 Wen where Wen is weight of each link e
or node n in path pi and can be represented as;
W=ae+ 1
bn
where ae represents the load of link e
whereas, bn is the battery level of each node n.
Notably, a higher value of battery has a less
weighted value. One major reason to use this
46
fraction is to handle the overall weight values of
different paths. For example, if two paths have
similar link load but significantly different battery
levels or vice versa, then the overall weight
value can help to select the best path. Formally,
the problem can be defined as; a source vs
intends to communicate with vt, where vs, vt 2 V
and path pi is said to be best optimal path if: Wpi
Wpj
, where pi, pj2P^ pi 6¼ pj.
The problem statement can be presented with
the help of sample topology, as shown in Figure 1.
These UAVs with different coverage areas and battery
levels are connected using a data link to form
a UAVNet. The SDN-based control center is associated
with these UAVs using a control link. In addition,
the UAVNet has multiple paths between the
source and destination. SDN-based UAVNets have
four components: UAVs, consumers, UAV control
center, and SDN controller. UAVs are equipped
with WiFi access points (APs) with different transmission
ranges and provide various services to
consumers in the sky. These consumers are terminal
devices (e.g., mobile nodes or sensors) distributed
unevenly on the ground using UAVs to
transfer their data. The position of UAVs can be
controlled with the help of a UAV control center,
which also has informationabout the current location
and battery level of UAVs.
SDN controller communicates with all UAVs
coming under its domain using OpenFlow protocol
to receive packets from UAVs. It also extracts
topology updates, link states, and statistics (e.g.,
latency, packet loss ratio, and link utilization).
After collecting this information, the controller
can forward it to management plane applications.
Since there can be multiple paths between
source and destination, for example, let UAV1
and UAV6 are the source and destination nodes
and there are two possible routes among this
pair, which can be represented as path P1 and P2
(i.e., UAV1 !UAV2 !UAV4 !UAV6 and UAV1
! UAV3 ! UAV5 ! UAV6), as shown in
Figure 1.
As soon as UAV1 receives a packet, it forwards
the packet to the controller as Packet_IN
message. In response to the Packet_IN request,
the controller generates a Packet_OUT message,
which contains flow rules for each UAV of path
P1. After the flow rule installation, the same path
is followed for the flow duration. This situation
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023
