EFFICIENT DATA GATHERING SCHEMES FOR WIRELESS SENSOR NODES

INTRODUCTION
WSNs and major applications[2,3] Our area of focus
Efficient schemes for data gathering
Stationary nodes [4,5,6] Mobile nodes Triangulation Algorithm [1]

TRIANGULATION ALGORITHM [1]

MAIN OBJECTIVES
Cover the Entire Area Minimize revisits Optimize a certain parameter in the process
Distance Time A little bit of both

DISTANCE OPTIMIZATION[1]

Row wise and Col wise moves to optimize distance Causes time taken to increase

TOTAL DISTANCE COVERED

Dt= C1Nrow + C2Ncol1+ C3Ncol2 + C4Nrev Where o C1 = Co-eff. for row wise movements Nrow o C2 = Co-eff. for type 1 col wise movements Ncol1 o C3 = Co-eff. for type 2 col wise movements Ncol2 o C4 = Co-eff. for revisit movements Nrev
(Co-effs. may be r3 or 2r3)

TOTAL TIME TAKEN

Tt = t3[Nrow+ Ncol1 + Nrev] + t3Ncol2 = t3[Nrow + Ncol1 + Ncol2] + t3Nrev Where * time t3 to cover a distance of r3

*We assume that to cover a distance r we take time t. Hence to cover a distance r3 node will take t3 time

BOUNDS ON DISTANCE AND TIME

Dt= C1Nrow + C2Ncol1+ C3Ncol2 + C4Nrev [1]
C1 = C4 = r3 and C2, C3 may be r3 and 2r3

Tt = = t3[Nrow + Ncol1 + Ncol2] + t3Nrev Lower Bound - Max type 1 column wise moves (r3) - No Revisits Upper Bound - Max type 2 column wise moves (2r3) - Revisits

TIME OPTIMIZATION[1]

Row wise and Col wise moves to optimize time Key is to make 2 node movements at once each time Causes distance travelled to increase substantially

TOTAL DISTANCE COVERED

Dt= C1Nrow + C2Ncol1+ C3Ncol2 + C4Nrev Where o C1 = Co-eff. for row wise movements Nrow o C2 = Co-eff. for type 1 col wise movements Ncol1 o C3 = Co-eff. for type 2 col wise movements Ncol2 o C4 = Co-eff. for revisit movements Nrev
(Co-effs. may be 2r or 2r3)

TOTAL TIME TAKEN

Tt = t[Nrow+ Ncol1 + Nrev] + t3Ncol2 Where time t to cover a distance of r and t3 to cover a distance of r3

BOUNDS ON DISTANCE AND TIME

Dt= C1Nrow + C2Ncol1+ C3Ncol2 + C4Nrev Tt = t[Nrow+ Ncol1 + Nrev] + t3Ncol2

Lower Bound - Max type 1 column wise moves (2r) - No Revisits Upper Bound - Max type 2 column wise moves (2r3) - Revisits

DISTANCE AND TIME OPTIMIZATION

Equal weight to both time optimization as well as distance optimization but can be extended to unequal weights as well Movements are alternately distance optimized and time optimized

TOTAL DISTANCE
Dt= C1Nr-d + C2Nr-t + C3Nc-a + C4Nc-b + C5Nc-c + C6Nr-r-d + C7Nr-r-t
Where Nr-d = Row wise movements which were distance optimized, co efficient of r3 Nr-t = Row wise movements which were time optimized, co efficient of 2r Nc-a = Column wise movements with co-eff. 2r3 Nc-b = Column wise movements with co-eff. r3 Nc-c = Column wise movements with co-eff. 2r Nr-r-d = Row wise revisits which were distance optimized, co-eff. r3 Nr-r-t = Row wise revisits which were time optimized, co-eff. 2r

TOTAL TIME
Tt= C1Nr-d + C2Nr-t + C3Nc-a + C4Nc-b + C5Nc-c + C6Nr-rd + C7Nr-r-t
Where Nr-d = Row wise movements which were distance optimized, co efficient of t3 Nr-t = Row wise movements which were time optimized, co efficient of t Nc-a = Column wise movements with co-eff. t3 Nc-b = Column wise movements with co-eff. t3 Nc-c = Column wise movements with co-eff. t Nr-r-d = Row wise revisits which were distance optimized, co-eff. t3 Nr-r-t = Row wise revisits which were time optimized, co-eff. t

BOUNDS ON DISTANCE AND TIME

Lower Bound Dlb= r3ceil(2(Ymax 1) + (Xmax 2)(Ymax 2)) +2rfloor(2(Ymax 1) + (Xmax 2)(Ymax 2)) + r3( (Xmax-1)/2 + (Xmax-2) +1) +2r3((Xmax-3)/2)

Tlb= t3ceil(2(Ymax 1) + (Xmax 2)(Ymax 2)) +t floor(2(Ymax 1) + (Xmax 2)(Ymax 2)) + t3( (Xmax-1)/2 + (Xmax - 2) +1) +t3((Xmax-3)/2)

BOUNDS ON DISTANCE AND TIME

Upper Bound Dub = 2rceil(2(Ymax 1) + (Xmax 2)(Ymax 2)) +r3floor(2(Ymax 1) + (Xmax 2)(Ymax 2)) + 2r3((Xmax-1)/2 + (Xmax-2) +1) + r3(Xmax-3)/2 Tub = t ceil(2(Ymax 1) + (Xmax 2)(Ymax 2)) +t3floor(2(Ymax 1) + (Xmax 2)(Ymax 2)) + t3((Xmax-1)/2 + (Xmax-2) +1) + t3(Xmax-3)/2

SIMULATION
A field of area 4500 * 2000 is chosen The communication radius = 50 Hence the area is divided into 47 * 181 triangles.

RESULTS DISTANCE TRAVELLED I

RESULTS DISTANCE TRAVELLED II

RESULTS DISTANCE TRAVELLED III

RESULTS TIME I

RESULTS TIME II

RESULTS TIME III

RESULTS INDIVIDUAL NODE DIST I

RESULTS INDIVIDUAL NODE DIST II

RESULTS INDIVIDUAL NODE DIST III

RESULTS NODE DEVIATION

COMPARING RESULTS

DATA GATHERING SCHEME WHEN COVERAGE AREA HAS A HOLE

Problems with the current scheme How are we going to solve it ?

THE ALGORITHM
(1,1) (1,Ymax)

Snake Like Traversal Pattern Row Wise Column Wise Any Block can be covered first Avoid Revisits
(Xmax,1) (Xmax,Ymax)

DERIVATION OF BOUNDS- DISTANCE

Lower Bound Max moves should be row wise moves. Row Wise Moves= (Ymax-1)(Xmax)*r 3 Column Wise Moves= ((Xmax-1/2)) r 3 + (Xmax-1/2)2 r 3 Moves lost due to hole= Nholer 3 Final Expression= Row wise + Column Wise Lost Moves

DERIVATION OF BOUNDS- DISTANCE

Upper Bound Max moves should be column wise moves. Col. Wise Moves=((Xmax-1)*r 3 +(Xmax-1)*2r3)* (Ymax-1/2) +(Xmax-1)/2*r 3 +((Xmax-1)/2 )2r 3 Row Wise Moves= (Ymax-1)*r 3 Moves Lost due to hole= Nhole/2[r 3 + 2r 3] Final Expression= Column Wise+ Row Wise- Lost Moves

DERIVATION OF BOUNDS-TIME
Lower Bound DLB= (Ymax-1)(Xmax)*r 3 +((Xmax-1/2)) r 3 + (Xmax-1/2)2 r 3 - Nholer 3 TLB= (Ymax-1)(Xmax)*t 3 +((Xmax-1/2)) t 3 + (Xmax -1/2)2 t 3 Nholet 3

DERIVATION OF BOUNDS-TIME
Upper Bound DUB= ((Xmax-1)*r 3 +(Xmax-1)* 2 r3)(Ymax1/2) + (Xmax-1)/2*r3 + ((Xmax-1)/2 )2r 3 + (Ymax-1)*2r3 - Nhole[r3 + 2r 3] TUB= ((Xmax-1)*t 3 +(Xmax-1)* t3)(Ymax- 1/2) + (Xmax-1)/2*t 3 + ((Xmax-1)/2 )t 3 + (Ymax-1)*t 3 - Nhole[t3 + t 3]

RESULTS
A 4500 *2000 field was chosen. This gives us 47*181 equilateral triangles The position of the hole is between triangle number 4,4 and 7,7 The algorithm is run 100 times

Total Distance
760000

755000

750000

745000

740000

735000

730000

725000 1 3 5 7 9 111315171921232527293133353739414345474951535557596163656769717375777981838587899193959799

Total Time
15500

15300

15100

14900

14700

14500 1 6 12 18 24 31 38 43 49 54 60 66 74 81 89 94 100

Deviation
6

0 1 4 8 11 14 18 22 25 29 35 38 41 44 48 51 54 58 62 65 70 74 79 84 88 91 94 98

PRIORITY BASED MOBILE TRAVERSAL ALGORITHM

Time Optimized Algorithms were covered in the first part of this talk, BUT.

THE ALGORITHM
A (30%) B (20%)

D (20%)

C (30%)

RESULTS
1600 1400 1200 1000 800 600 400 200 0 -200

Difference in Time

In Terms of Total Time Taken to find the Object, Priority Based MTA Wins Big Time

500

400

300

200

Difference in Distance Covered

100

-100

-200

In Terms of Total Distance Covered to find the Object, Priority Based MTA loses

RECAP
Distance Optimized MTA Time Optimized MTA Weighted Average of Distance Time Optimization MTA MTA when coverage area has a hole Priority Based MTA

FUTURE WORK
Dimensions for the hole are not known in advance Improvement to the priority based MTA

REFERENCES
1. A. Khan, C. Qiao, S. Tripathi, Mobile traversal schemes based on triangulation coverage, In Mobile Networks and Applications Volume 12, Issue 5 (December 12), Pages 442-437 Rmer, Kay; F. Mattern (December 2004). "The Design Space of Wireless Sensor Networks". IEEE Wireless Communications 11 (6): 54- 61. doi:10.1109/MWC.2004.1368897 Thomas Haenselmann (2006-04-05). Sensor networks. GFDL Wireless Sensor Network textbook. Cardei M, Thai M, Li Y, Wu W, Energy-efficient target coverage in wireless sensor networks. In IEEE INFOCOM. Miami, 1317 March 2005 Carle J, Simplot D, Energy efficient area monitoring by sensor networks, IEEE Comput 37(2):4046, 2004 Tian D, Georgannas N (2002) A coverage-preserving node scheduling scheme for large wireless sensor networks. In: ACM workshop on WSNA. Atlanta, 28 September 2002

2.

3. 4.

5. 6.

QUESTIONS AND COMMENTS

THANK YOU!