6.3 Link Quality Assessment
6.4.1 LQA experiments
The first set of experiments we conducted aimed at defining an efficient LQA and verifying its accuracy against the actual data packet delivery ratio. Fig- ure 6.4 shows the same degrading scenario as in Figure 6.3, but with different LQA mechanisms. Two laptops were used, where one slowly moved away from the other. The fixed laptop generated 50 data packets of 1000 Bytes per second at a fixed transmission rate of 54 Mbps. At the same time, both laptops generated one hello packet per second of 80 Bytes using broadcasting at a transmission rate of 2 Mbps. This is the most natural setup for multi- hop ad hoc networks based on IEEE 802.11 technology. Unfortunately, this frequently leads to performance problems when the delivery ratio of the hello packets is used as LQA as explained earlier in this chapter.
The black thick line in Figure 6.4 shows the delivery ratio of the hello packets using an EWMA with α = 0.2. Compared to the actual data delivery
0 0.2 0.4 0.6 0.8 1 -90 -80 -70 -60 -50 -40 Delivery Ratio Signal Strength 0 0.2 0.4 0.6 0.8 1 -90 -80 -70 -60 -50 -40 Delivery Ratio Signal Strength
(a) Received Signal Strength (b) Processed Signal Strength
Figure 6.5: Signal strength vs. data packet delivery ratio
ratio, illustrated by the gray background graph, it can be seen that the hello packet delivery ratio constantly overestimates the quality of the link. This is due to the smaller packet size and the slower transmission rate being used. Even when no data packets can be delivered, the hello packet delivery ratio indicates a delivery ratio in the range of 10 to 50 %. Furthermore, adaptation to link quality changes is slow. Hence, hello packets alone are not a good approach to LQA.
In Figure 6.4, the black thin line shows the received signal strengths of the hello packets with a lost packet assuming a signal strength of −95 dBm, which is the lowest receivable signal strength for our equipment. It can clearly be seen that the signal strength degrades as the two laptops move apart, but it can also be seen that the signal strength fluctuates. Hence, we need to smooth it out with EWMA, which is shown as the dark gray thick line (α = 0.2).
A better approach to LQA, when there are no or very little data packets on the link, is to use a combination of the hello packet delivery ratio and the received signal strength. To investigate the potential of such a combi- nation, we designed an experiment where one laptop constantly sent 1000 Bytes packets to another at a rate of 20 packets/s. The transmitting laptop was moved around to create different scenarios with different packet delivery ratios and different received signal strengths. Figure 6.5 shows the result of this experiment where each measure point contains the average over 1 second (around 20 packets). Figure 6.5(a) shows the average received signal strength versus the average delivery ratio. The correlation is not clear as the signal strength drops below −60 dBm. When we account for the lost packets by assuming a received signal strength of −95 dBm, we get the result of Fig- ure 6.5(b). Suddenly, the correlation becomes much stronger; this hints at a linear relation below the threshold of −60 dBm.
6.4. EXPERIMENTAL RESULTS 131
packet delivery ratio as a LQA, we used the following formula:
LQEST = MIN(1, C · (1 −
SSHR
SSM IN
) · LQHR) (6.2)
where LQHR is the EWMA of the packet delivery ratio, SSHR is the EWMA
of the received signal strength, SSM IN is the minimal receivable signal
strength, and C is a tunable parameter. There are other ways of defining an estimation function based on hello packet delivery ratio and received signal strength, but this approach was selected in order to identify the applicability of using the combination of hello packet delivery ratio and received signal strength as LQA. Further research may determine if a different formula is better.
In some experiments presented in [242][243], we identified the optimal value of C, which actually depends on the data rate and the packet size. Since we used a rate of 54 Mbps and data packets of 1000 Bytes, the best C was found to be 2.6 with α = 0.2 for the EWMA of both the hello packet delivery ratio and the received signal strength. Only signal strength values of hello packets were used. It should also be noticed that these parameters still work well with other packet sizes and transmission rates.
We again used the same setup as in the degrading link experiments in Figure 6.3 and Figure 6.4, but this time with a different scenario. The mobile laptop moved to and from the fixed laptop with each cycle taking about 40 seconds. During each cycle, the link turned from perfect to non-existent and then back to perfect again. The gray background graph in Figure 6.6 shows the data delivery ratio as in the previous figures. The thick black line shows the LQA based on only the hello packet delivery ratio, while the thin black line shows the LQA based on formula (6.2).
LQEST based on formula (6.2) is far from a perfect LQA, but is certainly
a big improvement compared to only using the hello packet delivery ratio. In the following section, we will investigate how this LQA works in a small three-node network.