IEEE Consumer Electronics Magazine - January/February 2023 - 49

Figure 3. RTT with and without using load
balancing algorithm.
Figure 4. Percentage of packets processed on
each node.
rate, distance, and packet transmission rates.
Due to flight, the battery of each UAV is reduced
at a fixed rate. Since transmission power is
directly proportional to distance, higher distance
consumes more battery. Similarly, battery life
relates to the number of packets transmitted;
thus, more battery is consumed if any UAV transmits
more packets. Each UAV is connected to one
or more UEs (stations). Iperfc generates traffic
between the source and destination pair. To
show the effectiveness of the proposed
approach, a comparison with the state-of-the-art
LB-UAVnet is presented. In addition, the proposed
approach is compared with the default
controller decisions (i.e., without applying the
proposed approach).
In this article, one of the significant parameters
for evaluation is round trip time (RTT), as
shown in Figure 3, where the green axis represents
minimum, average, and maximum RTT
without applying load balancing algorithm (w/o
LBA). In contrast, the blue axis shows minimum,
average, and maximum RTT when the load balancing
algorithm (w LBA) is applied. The upper
value indicates the maximum, the lower value
shows the minimum, and the middle value represents
the average RTT. In each experiment, five
packets are transmitted between the source and
destination pair. It can be observed that the minimum
RTT in both cases is almost similar,
whereas there is a significant difference in maximum
RTT. The maximum RTT w/o LBA is more
than 6 ms in all experiments, whereas, in experiment
3, this value is 8 ms. On the other hand, w
LBA is applied, this value does not exceed
c
[Online]. Available: https://iperf.fr/
4.9 ms. Similarly, the average RTT is above 5 ms
when LBA is not applied, but it is reduced to
4.84 ms after applying the algorithm.
Figure 4 shows the percentage of packets processedoneachUAVworkingas
APs.The percentage
of packets processed on each UAV w/o LBA is indicated
as a green bar whereas the red bar shows LBUAVnet,14
and the blue bar showswLBA is applied.
Moreover, the average values of each approach are
presented in the dashed lines. It can be observed
that w/o LBA the path selected is UAV1 ! UAV3
! UAV5 ! UAV6 (i.e., Path 2, as shown in
Figure 1). It can be seen that the percentage of
packet processing is more than 60% on these
UAVs. On the other hand, packet processing over
UAVs of other path (that includes UAV2 and
UAV4) is almost 0%. Notably, the average value
without LBA is computed by considering the UAVs
processing the packets, whereas the UAVs with
zero workloads (i.e., UAV2 and UAV4)are
neglected. Similarly, the packet processing on each
UAV with LB-UAVnet is between 35% and 40%. On
the other hand, it can be evident that when the proposed
approach is applied, the packet processing
of UAV2 and UAV4 is also increased to 30%
whereas load on UAV3 and UAV5 is reduced. The
packet processing after applying LBA is between
30% and 35%. It indicates that the proposed
approach distributes load evenly on multiple
paths. Furthermore, the proposed approach outperforms
LB-UAVnet as well. One possible reason
is that LB-UAVnet balances the load based on flow
rules rather than network traffic.
To compute link utilization on all possible
paths from source to destination without battery
considerations, two applications are developed.
One application generates the observed data flow,
January/February 2023
49
https://www.iperf.fr/
IEEE Consumer Electronics Magazine - January/February 2023

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