Presentation is loading. Please wait.

Presentation is loading. Please wait.

Fairness Matters: Identification of Active RFID Tags with Statistically Guaranteed Fairness Michigan State University Muhammad Shahzad Alex X. Liu North.

Similar presentations


Presentation on theme: "Fairness Matters: Identification of Active RFID Tags with Statistically Guaranteed Fairness Michigan State University Muhammad Shahzad Alex X. Liu North."— Presentation transcript:

1 Fairness Matters: Identification of Active RFID Tags with Statistically Guaranteed Fairness Michigan State University Muhammad Shahzad Alex X. Liu North Carolina State University

2 2 Radio Frequency Identification ActivePassive

3 3 Active RFID Tags RailwaysSeismology AutomotiveAircraft

4 4 Tree Walking 0 00 000 001 01 1 10 11 010 011 100101 1000100110101011 1 2 3 4 5 67 8 9 10 11 12 13 1415 16

5 5 Problem Statement  Input ─ RFID tag population of unknown size ─ Required fairness = α  Output ─ IDs of all tags ─ Minimize identification time ─ Achieved fairness ≥ α

6 6 Interpreting Fairness  Example: ─ Tag battery depletes after 100,000 transmissions ─ One thousand identification rounds per day ─ Fairness = 0.84 ● 20% tags last for 33 days ● 30% tags last for 50 days ● 50% tags last for 100 days ─ Fairness = 0.99 ● 1% tags last for 50 days ● 99% tags last for 100 days

7 7 Communication Protocol Overview esscess 326447 1234567 esscess Frame size f i = 7 Number of empty slots: e i Number of successful slots: s i Number of collision slots: c i the tunable parameter

8 8 Proposed Approach 1.Estimate tag population size ─ Using ART [MobiCom 2012] ─ One time cost 2.Calculate optimal frame size and execute frame 3.Re-estimate unidentified tag population size ─ Go to step 2 optimal frame size

9 9 Jain’s Fairness Index  x l = amount of resource used by l th node  t = total number of nodes  Jain’s fairness index lies in the range [1/t, 1]

10 10 Jain’s Fairness Index for Tags  x l = number of times a tag with label l transmits  t = total number of tags

11 11 Jain’s Fairness Index for Tags where  Smaller the load factor, higher the fairness  Trade-off time for fairness load factor

12 12 Identification Time

13 13 Constraint Optimization Problem  Now we know load factor k  We already know t i from re-estimation  We get f i, because k = t i /f i Trading-off time for fairness Just optimal Aloha

14 14 Handling Large Frame Sizes  Max allowed frame size = f max  Divide population into 2 z groups of equal size ─ where, z = ceil ( log 2 { f i / f max } )  Execute frames of size ceil( f i / 2 z )  Use SELECT command to make t / 2 z tags participate for each frame ─ LSBs of tags are almost uniformly distributed ● Tags with IDs ending in 0 = Tags with IDs ending in 1 ─ Example: to divide into four groups 1.Use SELECT with 00 2.Use SELECT with 01 3.Use SELECT with 10 4.Use SELECT with 11  Proof of fairness for this method in paper

15 15 Experimental Evaluation  Implemented 9 protocols in addition to FRIP 1.BS (IEEE Trans. on Information Theory, 1979) 2.ABS (MobiHoc, 2006) 3.TW (DIAL-M 2000) 4.ATW (Tanenbaum, 2002) 5.STT (Infocom, 2009) 6.MAS (PerCom, 2007) 7.ASAP (ICDCS 2010) 8.Frame Slotted Aloha (IEEE Transactions on Communications, 2005) 9.TH (MobiCom 2012)

16 16 Fairness

17 17 Identification Time

18 18 Effect of Splitting Tag Population

19 19 Comparison: Fairness

20 20 Comparison: Time

21 21 Conclusion  First effort towards developing a fair RFID identification protocol  Proposed a method to achieve the required fairness while minimizing identification time  More in the paper ─ Formal proofs and derivations of various aspects ─ More comparisons of FRIP with prior protocols

22


Download ppt "Fairness Matters: Identification of Active RFID Tags with Statistically Guaranteed Fairness Michigan State University Muhammad Shahzad Alex X. Liu North."

Similar presentations


Ads by Google