1 / 24

A hybrid randomized protocol for RFID tag identification

A hybrid randomized protocol for RFID tag identification. Aleksandar Micic Amiya Nayak David Simplot-Ryl Ivan Stojmenovic. RFID. R adio F reqency ID entification Technology for tracking objects Reader device sends probes to a set of RFID tags

shandra
Download Presentation

A hybrid randomized protocol for RFID tag identification

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A hybrid randomized protocol for RFID tag identification Aleksandar Micic Amiya Nayak David Simplot-Ryl Ivan Stojmenovic

  2. RFID • Radio FreqencyIDentification • Technology for tracking objects • Reader device sends probes to a set of RFID tags • Only reader has signal/collision detection capability • A tag is recognized only when it is the only one to respond (status=single) • If none of tags responds then status=silence • If two or more tags respond then status=collision

  3. Tag Identification Problem • Minimize the number of probes necessary for reading all the tags • Assuming that the number of tags is known in advance (to simplify) • Assuming single channel for all communication • Time divided, slots divided in two sub-slots: probes from reader and responses by tags • Reader informs about the outcome: silence, collision or unique message at the beginning of the next slot as part of new probe

  4. Related work • Hybrid randomized protocols for batch conflict resolution and estimation n of the number of stations • Cidon, Sidi (1988): 2.43n for small n to 2.13n for large n, • Greenberg and Ladner (1987) • Popovski, Fitzek, Prasad (2004): 2.35n to 2.07n. • Our protocol is similar in flavor but details are different; it is simpler due to assumptions of known n and we provide elegant proof for the bound

  5. Optimizing probability • Suppose that there are m tags, and each responds with probability p. • What is the expected number of tags to respond? • What value of p will optimize the probability that a single response is received?

  6. Identification by optimizing probability Assign IDs to stations or read tags: Protocol n-partitionHayashi, Nakano, Olariu 1999 form  ndownto 1 do { repeat station broadcasts on the channel with probability 1/m until the status on the channel is single broadcast the tag that has broadcast is read and stops competing for ID } Approximately en competing steps, e=2.71..

  7. {a,b,c,d,e} {a,c,d} {b,e} {a,d} {e} {c} {b} {a} {d} Partitioning for initialization HNO’99 repeat • each tag selects 0 or 1 with probability 1/2; • all the tags that selected 1 broadcast; • let Status1 be the resulting status; • all the tags that selected 0 broadcast; • let Status0 be the resulting status; • until Status1 <> silence and Status0 <> silence; Successful partitions

  8. Partitioning algorithm HNO’99 L= # of sets to identify tags from P(1),…,P(L)= sets of tags to identify Tag in P(k) participates in identification if k=L, otherwise it waits Drawbacks: If status1=silence then status0 known without probe Calculating |P(L)| does not need extra step not correct for 1 tag 2-partition L  n  1; while L >= 1 do { if |P(L)| = 1 then { n-th tag read; n  n + 1; L  L – 1 } else { partition P(L) into P(L), P(L+1); L  L + 1; };};

  9. Our modifications • If status1=silence then status0 not asked (if the first part is empty then there must be something in the second part) • Very first transmission not necessary since set of tags is known not to be empty • Introduced one extra time slots after any unique transmission, which will ask all competing tags to transmit • Goal: improve performance for small n=2,3,4, worsen performance for large n • It will be used in hybrid algorithm only for small n

  10. Protocol 2-partition-1 k  L NextId 1; station active; Repeattag with k < L just listens and updates NextId and L; repeat tag selects 0 or 1 with probability 1/2; active tag that selected 1 broadcasts  Status1 until Status1 <> silence; if Status1=single then { NextIdNextId + 1 }; active tag that selected 0 broadcasts  Status0; if Status0 = single then { NextIdNextId + 1}; if Status1  collision and Status0  collision then { L  L - 1}; if Status1 = collision and Status0 = collision then { L  L + 1; station in partition1: k  L } until L=0;

  11. Performance • How this works if there are no tags? • How this works if there is only one tag? How many probes? • Suppose that there are two tags. • F(2)= expected number of probes with 2 tags; • Possible outcomes in the internal ‘repeat’ loop are: • 00 (status1=silence), 01 (status1=single), 10 (status1=single), 11 (status1=collision). • probability of non-silence is p=3/4.

  12. Two tags • If probability of an event is p, the the expected number of attempts for event to happen is 1/p. • It takes 4/3 attempts to exit internal ‘repeat’ loop • At exit, there are three options: 10, 10, 11 • Probability of single is 2/3, and in one more probe the other tag is read • Probability of collision is 1/3, and it leads to the external ‘repeat’ loop, and F(2) probes • F(2)=4/3+2/3*1+1/3*F(2) -> 2/3*F(2)=2 • F(2)=3 probes

  13. Three tags F(3) probes • Outcomes external ‘repeat’: 000, 001, 010, 100, 110, 101, 011, 111 • Probability non-silence is 7/8, with 8/7 attempts av. • Status1=single for 001, 010, 100, 3/7 prob., will read collision of two other tags in the next probe, and will go back to external ‘repeat’ loop with F(2) probes; 3/7*[1+F(2)] • Status1=collision for 111, 1/7 prob., will probe status0 for silence and to external ‘repeat’: 1/7*[1+F(3)] • Status1=collision for 011, 101, 110; 3/7 prob., will read single with the next probe, and go back to the external ‘repeat’ loop with F(2) probes; 3/7*[1+F(2) • F(3)=8/7+3/7*[1+F(2)]+1/7*[1+F(3)]+3/7*[1+F(2)]; F(2)=3. • 6F(3)=8+12+1+12=33; F(3)=5.5

  14. Four tags • 0000: 16/15 attempts not to get it • 0001, 0010, 0100, 1000: 4/15*[1+F(3)] • 0011, 0101, 0110, 1001, 1010, 1100: 6/15*[1+2F(2)] • 1110, 1101, 1011, 0111: 4/15*[1+F(3)] • 1111: 1/15*(1+F(4)) • F(4)= 16/15+4/15*[1+F(3)]+6/15*[1+2F(2)]+4/15*[1+F(3)]+1/15*[1+F(4)] • 15F(4)=16+4*6.5+6*7+4*6.5+1 • F(4)=111/15=7.4

  15. Performance of 2-partition-1 • Two tags: expected 3 slots per tag: improved from expected 5 slots • Three tags: expected 5.5 slots • Four tags: 7.4 slots • For large n it is (e+1) per tag

  16. Hybrid initialization protocol • Binary partition for large n is normally successful • Use k-ary partitioning instead of 2-partition k=? • Repeat each active tag transmits with probability k/n {perform k-partitioning k=p/n p=?} casesilence: do nothing unique transmission: read tag, reduce n (# of tags) collision: call 2-partition-1 for collided tags; update n Endcase until all tags are read - binary (=2-ary) partition p=n/2 - p1 simple n-partition protocol • n dynamically updated if unknown

  17. Performance • Theorem 1. • The expected number of time slots in hybrid(n) protocol for reading of n tags is <2.20n. • Experiments: • 2.15n

  18. Conclusion • An extension for unknown number of stations described in master thesis by the first author • Tags have no memory capacity: (Kalinowski, Latteux, Simplot 2001) • Minimize radiation (# of bits sent by reader)? (Hush, Wood 1998; two patents in 1996)

  19. Unknown tag detection Liu, Qi, Li, Stojmenovic, Shen, Liu, Qu, 2014

  20. Basic principle: Bloom filter

  21. BF-UTD protocol • Filter length f and # of hash functions k to choose so that probability of detecting unknown tag existence is at least a • Identify verification: • hash N known tags, • broadcast filter F • Unknown tag reporting: • compare with F to decide if unknown, • if so send 1 bit; • if at least one bit sent, unknown tag presence (one or more) detected

  22. Best value for k ?

  23. Best value for k ?

  24. Enhanced algorithm • Sampling probability p for (known and unknown) tags to participate • Run same algorithm only for sampled tags • Optimal p and k depend on tolerance m (when m or more unknown tags are present), N, f, a

More Related