Experiment Results - Performance Evaluation and Results

Performance Evaluation and Results

5.2 Experiment Results

The first experiment was performed with constant radius of readers and variant number of RFID tags. The number of redundant readers can be detected in each algorithm is reported. Figure 12 shows LEO outperforms RRE in terms of the redundant reader detected. However, both techniques detect less number of redundant readers than the composite approach. As mentioned in last chapter, RRE+LEO scheme is not recommended in practice due to its very high algorithm overheads. Therefore, our experiments only show the results of LEO+RRE.

Figure 12: Comparison of redundant reader detected with network area 1000010000, reader radius=500 and number of reader=500.

Figure 13 shows comparison of algorithm overheads in different schemes. Algorithm overhead is referred as number of write operation issued by all readers in RFID network in order to perform redundant reader identification. We observe that algorithm overheads has the order LEO < LEO+RRE < RRE. This phenomenon matches our expectation. LEO has least number of write operations and RRE performs worst. A simple reason for this result is because LEO has (m) as upper bound of write operation while RRE has (2m) as lower bound if there are m RFID tags in the network.

Furthermore, if an RFID tag is covered by r readers on average, RRE will have (2mr) as lower bound of write operation while LEO remains (m) as upper bound. Reason for LEO+RRE has less algorithm overheads than RRE is because LEO uses the same memory space Rid and removes most of redundant readers in the first algorithm phase, the overheads of RRE algorithm could be largely reduced in the second phase. The LEO+RRE will have (2m(r-d)) as lower bound of write operation, if there are d RFID readers removed by LEO.

110 130 150 170 190 210 230 250 270 290 310

1000 2000 3000 4000 5000 6000 7000 8000 9000

Tags

Number of write

RRE LEO LEO+RRE RRE+LEO

0 50 100 150 200 250 300 350 400 450

500 600 700 800 900 1000

Radius

Redundants Reader Detection

RRE LEO LEO+RRE RRE+LEO

Figure 14: Comparison of redundant reader detected with network area 1000010000, number of tags=4000 and number of reader=500

For algorithm overheads, Figure 15 demonstrates the order LEO < LEO+RRE < RRE, which is similar to the observation we obtained in Figure 13. We also observe that LEO almost has a constant number (equal to number of tags) of write operation even under different reader radius. This is because the number of tags remains fixed in this experiment. As we analyzed before, LEO+RRE scheme has mean performance while the RRE performs worst in terms of algorithm overheads. Note that if we added RRE+LEO scheme for comparison, we will have the order LEO < LEO+RRE < RRE <

RRE+LEO for algorithm overheads and have the order LEO+RRE  RRE+LEO > LEO

> RRE for algorithm efficiency. This result encourages that LEO or LEO+RRE is most suitable for the redundant reader problem.

Figure 15: Comparison of number of write operations with network area 1000010000, number of tags=4000 and number of reader=500

Figure 16: shows the performance comparison with constant number of tags and variant phase scheme. It is perform LEO or RRE algorithm in multiple phase scheme in order to obtain the maximum redundant reader detection. As shown in Figures 16(a) and (b), the LEO algorithm has superior performance in term of larger number of redundant reader accumulated and detected.

100 125 150 175 200 225 250 275 300

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Phases

Accumulated Redundant Reader

RRE LEO

(a)

0 5000 10000 15000 20000 25000 30000 35000

500 600 700 800 900 1000

Radius

Number of Write

RRE LEO LEO+RRE RRE+LEO

0 20 40 60 80 100 120 140 160 180

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Phases

Redundant reader detected

RRE LEO

(b)

Figure 16: Comparison of number of redundant reader detected (a) and accumulated (b) with network area 1000010000, reader radius=500, number of tags=4000, reader=500

According to the above experimental results, we also obtained the RRE has lower performance of redundant reader detected and accumulated, even changed the sequence of query of the reader. We will illustrate base on Figure 17.

Figure 17: Fourth example of wireless RFID network with redundant reader R1

R3 R7

T1 T2

T4 T5

T7 T8

T10 T11

Figure 17 shows an RFID network contains seven readers, R1-R7 and eleven tags, T1-T11. It apt to find out R6 and R7 are redundant readers, but there are no redundant reader could be detected by RRE algorithm. We analyze the reasons and illustrate as following. Every RFID reader R1,R2,R3,R4 are covered 2 tags respectively, R5 is also covered 3 tags, and each reader covered at least one tag which is not cover by other reader. Therefore, R1-R5 could not be turned off by RRE algorithm. Beside, R6 and R7 are also covered T2, T6, T7 and T8, T9, T10 respectively; and that these six tags are also covered by R1-R5. Due to the tag count of reader R6 and R7 are more than R1-R5, therefore the tags T2, T6, T7 will be covered by reader R6 and T8, T9, T10 will be covered by reader R7. After finishing RRE scheme, these tags will be held by reader respectively. That means no redundant reader could be detected, even changed the sequence of query of the reader. Finally, we obtained a conclusion; some redundant reader might not be detected by RRE algorithm because these redundant readers have the maximum number of tag count.

CHAPTER 6

在文檔中中華大學 (頁 31-38)