RFID Readers Networks: A network layer perspective

Post on 23-Feb-2016

45 views 0 download

Tags:

description

RFID Readers Networks: A network layer perspective. Ahmed Jedda. Outline. RFID Readers networks Problems: Readers to tags collisions Tags discovery Network planning algorithms RFID coverage Readers connectivity. WSN vs. RFID . - PowerPoint PPT Presentation

transcript

RFID Readers Networks: A network layer perspective

Ahmed Jedda

Wisense workshop 2

Outline• RFID Readers networks• Problems: – Readers to tags collisions– Tags discovery– Network planning algorithms– RFID coverage – Readers connectivity

Wisense workshop 3

WSN vs. RFID • RFID offers much cheaper techniques

when it comes to identification of physical objects

– Similar to the radar problem found in the WWII

– No need to guess your object,

Wisense workshop 4

What is an RFID ?An RFID system consists:

- RFID reader, - RFID tag,

Procedure: - RFID reader sends a signal to an RFID tag, - An embedded circuit in the tag is

energized, - A reply is sent back to the reader

- The reply may have the identifier of the tag,

Notes: 1. No battery in RFID tags, (i.e. communication initiated by readers)2. Simple computation is allowed in tags, 3. RFID readers may write in the memory of neighbor tags,

Wisense workshop 5

Applications• Many applications

– Children temperature monitoring in Hongkong hospitals (integration with sensors)

– Assets monitoring: Indian TV station (assets worth millions of dollars)

– Animal tracking (cows tracking in of the oldest applications of RFID)

– Cars in parking (automatic payment)

• Technology is sponsored by major companies with world-wide supply chains (Walmart, Gillete, DHL, etc ..)

Wisense workshop 6

RFID readers networks models

Typical scenarios1 reader --- 1 tag,

1 reader ---- multiple tags multiple readers ----- multiple tags

Readers can interact with tags, Reader can interact with readers,

=RFID reader networks !

As the number of readers and tags increase, we get dense RFID readers

networks and we get many other problems as

well !

1 reader, 1 tag

1 reader, many tags

many readers, many tags

Communication between readers is possible

The readers collision problem

Wisense workshop 8

The readers collision problem

Problem: Two types of collisions may be caused by readers: 1. Readers to readers (a wireless networking problem)2. Readers to tags

Approaches:Assign the readers time slots to transmit in them such that no collision happen

- Implemented by: Colorwave

Wisense workshop 9

ColorwaveAlgorithm idea: 1. Create a graph G with vertices V = readers, E = the set of

interference edges (i.e. if two reader may cause interference to each other, then add

an edge between them in E.)

2. Color the vertices with colors {c1, …, ck}3. Sequence of time slots are given colors {c1, …, ck} 4. A reader with color ci transmits at the time slot with color ci

Colorwave: Random coloring - Each reader selects a random color, - If there is a collision, the reader and its neighbors select new colors, - to avoid collisions, set k to be high enough.

The tag discovery (identification) problem

Wisense workshop 11

The tag identification problem

Problem: A reader must find neighboring tags in the shortest possible time Basic solution: Broadcast and wait for replies. - The issue is the collisions caused by tags replying at the same timeApproaches: Aloha-based (based on backoff timers)Tree-based (similar to tree traversals - divide tags according to their id’s prefixes)

Wisense workshop 12

Aloha-based tags discoveryAssumptions: A tag have timer, can generate pseudo-random numbers, may send multiple replies.

Algorithm idea: 1. Readers broadcast inquiry messages, 2. Upon reception: 2.a: a tag backoff for a random time, 2.b: then, the tag sends back a reply3. A tag must receive an acknowledgement from the reader. 3.a: if received: be quite in the future. 3.b: otherwise, repeat the same previous procedures Improvements: •Synchronization of readers and tags,•Provide tags with time frame for better backoff periods, •Tags send only once in a frame

Random backoff

Wisense workshop 13

Tree-based tags discoveryAssumptions: A reader may inquire all neighbor tags with id’s prefix [p1,p2, …, pk].

Algorithm idea: • Send an inquiry message to tags with id = [p1,…,pk]XXX.

• Tags with id prefix [p1,…,pk] reply to inquiry. • Case1: no answer [no tags with such prefix is nearby]• Case2: one answer [only one node exists with such prefix]• Case3: collision [there are more than one with such prefix]

• In case of collision: • Generate pk+1 randomly – either 0 or 1• Repeat Steps with id [p1, ….pk+1]XXX until no answer, or one

exact answer !

This forms a binary tree

Wisense workshop 14

Tree-based tags discovery

Reader inquires all tagsId = [XXXX]

0010

1101

0100

0011

All tags reply, Collision is detected

0010

1101

0100

0011

Example

Wisense workshop 15

Tree-based tags discovery

Reader inquires all tagsId = [0XXX]

0010

1101

0100

0011

Tags {0010, 0100, 0011} reply, Collision is detected

0010

1101

0100

0011

Example

Wisense workshop 16

Tree-based tags discovery

Reader inquires all tagsId = [01XX]

0010

1101

0100

0011

Tag 0100 reply, Only one answer -- continue looking for other tags !

0010

1101

0100

0011

1

01

01

0

1

1101

0100

0

0011

0010

No answer

Example

RFID network planning

Wisense workshop 18

RFID Networks planningProblem: •What are the best positions for the readers in a network that let every tag covered by at least one reader? •Other performance measurements could be considered - minimize interference, balance load etc ..

Approaches:• Known-area topology: if the area is already known, then the

optimal placement of readers is possible such that every point is cover.• Examples: grid, honey grid.

• Heuristic search algorithms (Genetic algorithms, Swarm intelligence etc ..): • solve a set of objective formulas (usually complex ones) • Most of them assume previous knowledge of tags positions ! [what if

tags move then ?]• Set-cover based (3D)• Random placement

Wisense workshop 19

Set-cover based network planningAssumptions: • previous knowledge of tags locations, •Tags may exist in a cube, whereas readers only on borders. •Reading range is R

Observation:1. A tag centered at point (xt,yt,zt) can be covered by any reader located in a sphere centered at (xt,yt,zt) with radius R2. The intersection of the sphere with the walls is a circle, 3. Put readers in the tags circles intersection

Placing one reader will cover the tree tags

The problem is discrete now !

Solve it with a greedy approximation set-cover algorithm

T1T2

T3

RFID Coverage

Or simple, redundant readers elimination …

Wisense workshop 21

RFID Coverage

Problem: Given an RFID network such that every tag is covered, there may be more readers that what is required.

- Find the minimum amount of readers that cover every tag.

Solutions of the problem may be used to: 1) increase network lifetime, 2) fault-tolerance, 3) load balancing, and 4) data redundancy eliminationApproaches:Set-cover based,

- GREEDY, NTE Readers compete for tags ownership (assume tags are writable),

-RRE, GDE, LEO, RANDOM and others ..

Wisense workshop 22

First algorithm: RRERRE: (Algorithm idea):•Every reader writes (tag count, identifier) in the memory of all its neighboring tags. •Wait for all readers to write, •Read memory contents of neighbor tags, •A reader owns a tag if it has the largest (tag count, identifier).

(tag count, id) (1,1)

(3,2)

(2,3)

(1,4)

{(3,2), (1,1)}

{(3,2), (1,4)}

{(3,2), (2,3)}

{(2,3), (1,4)}

Non-redundant readers are t2 and t3

Other similar solutions: LEO: the first reader to write owns the tag,DRRE: the reader with maximum number of reader neighbors owns the tag

Wisense workshop 23

Improvement on RRE - GREEDY

GREEDY: (Algorithm idea):Let T = {t1, …., tm} : set of tagsLet N = {N(r1), N(r2) …., N(rn)} : the set of tags covered by a reader r, Find a subset N’ of N of minimum size such that the union of N’ is T.

Set-cover approximation algorithm:Always select the set N(ri) that minimize the number of uncovered tags.

Assumptions: centralized algorithm – a server exists

NTE is a similar algorithm, Maximum weight depends on tags count, and number of reader neighbors, Better than RRE, (but not GREEDY)

Wisense workshop 24

Improvements on RRE - GDE

GDE- Algorithm idea: 1. Runs in iterations, 2. A reader owns a tag only if it

owns every other active neighbor tags. (that is, if it has the largest (tag count, identifier))

3. Deactivate all owned tags, and repeat with new network until all tags are deactivated.

(2,3)

(2,2)

(1,1)

(tcount, id) (2,4)

(2,3)

(2,2)

(1,4)

(2,5)

Wisense workshop 25

Improvements on RRE - GDE

GDE- Algorithm idea: 1. Runs in iterations, 2. A reader owns a tag only if it

owns every other active neighbor tags. (that is, if it has the largest (tag count, identifier))

3. Deactivate all owned tags, and repeat with new network until all tags are deactivated.

(1,3)

(2,2)

(1,1)

(1,3)

(2,2)

(1,4)

Wisense workshop 26

Improvement on GDE – GDE-k

GDE-K Algorithm idea: 1. Run GDE for k iterations, 2. At the k+1 iteration, run RRE

Wisense workshop 27

Algorithm RANDOM+RANDOM+: Algorithm idea1. Each reader weight is (rand(r), id(r)), where rand(r) is a

random number selected by the reader at the beginning of an algorithm.

2. A tag is owned by a neighbor reader with maximum weight.

Benefits: Can be run multiple iterations:

- execution on non-redundant readers. - each iteration a new rand value is selected.

Random-6 (i.e. with 6 iterations) perform somehow similar (or even better sometimes) compared to GDE [with the same amount of write/read rounds]

We can also run RANDOM+ after GDE – (GDE-RANDOM+) not much improvement though !

Wisense workshop 28

Problems with writable tagsObservation:the read range Rr is larger than the write range Rw.

Problem:How to design algorithms that consider this case ?

Case: Rw < Rr: there is always a topology organization such that no tag is writable, Therefore: an algorithm should consider writing is not allowed (if Rw < Rr is to be considered).

Can we do RRE without writing ? [relax to tag reporting problem]-there must be links between readers, otherwise impossible. -readers should be connected - otherwise impossible. -If there are links, --> at least worse than election !

- that is, no localized algorithm - A lower bound: O(nlogn + m) in term message complexity.

In fact, RRE = directed graph construction GDE = topological sorting

Readers connectivity

Wisense workshop 30

How to connect all readers ?Problem:Find a connected multihop network that contains minimum

amount of readers to perform a certain task. In our case, task = coverage.

Assumptions:Readers can communicate with each other

Wisense workshop 31

How to connect all readers ?

Approaches: - Local rules: delete a node if it does’nt affect connectivity

(usually the k-hop neighborhood graph is considered) – - Minimum spanning trees- Shortest path trees: - Virtual trees- Connected dominating set

Problem:Given a graph G = (V,E) and a set V’ of V (called necessary nodes), find

an minimum sized V’’ (called relays) in V such that the induced graph of V’ and V’’ is connected ---- very similar to the minimum steiner tree problem

Wisense workshop 32

Minimum spanning tree based

Necessary nodesRelay nodesothers

Assume that this tree is the minimum spanning tree of graph G= (V,E)

Wisense workshop 33

Minimum spanning tree based

Necessary nodesRelay nodesothers

Leaves that are not necessary delete themselves

Wisense workshop 34

Minimum spanning tree based

Necessary nodesRelay nodesothers

Leaves that are not necessary delete themselves

Wisense workshop 35

shortest path tree basedAlgorithm idea:Start with a tree that T contains a single necessary node Find the closest necessary v node to the tree T, Add the nodes of the path between v and the tree T

Wisense workshop 36

shortest path tree basedAlgorithm idea:Start with a tree that T contains a single necessary node Find the closest necessary v node to the tree T, Add the nodes of the path between v and the tree T

Wisense workshop 37

shortest path tree basedAlgorithm idea:Start with a tree that T contains a single necessary node Find the closest necessary v node to the tree T, Add the nodes of the path between v and the tree T

Wisense workshop 38

shortest path tree basedAlgorithm idea:Start with a tree that T contains a single necessary node Find the closest necessary v node to the tree T, Add the nodes of the path between v and the tree T

Wisense workshop 39

shortest path tree basedAlgorithm idea:Start with a tree that T contains a single necessary node Find the closest necessary v node to the tree T, Add the nodes of the path between v and the tree T

Wisense workshop 40

Virtual treesAlgorithm idea:Assume knoweldge of tags positions, Build a tree of tags, This facilitates the tree construction process

Wisense workshop 41

Connected dominating sets

Algorithm idea:Use a localized algorithm to build a connected dominating

set, - we use the multiset relay algorithm

Then execute RRE, GDE or any other --- but give priority to nodes already in CDS.

Recall: A dominating set of a graph G = (V,E) is a set V’ such that every node in V is either in V’ or neighbor to a node in V’.A connected dominating set V’ restrict that the induced graph of V’ is connected

THANKS !

42Wisense workshop