Tracking Moving Devices with the Cricket Location System

Tracking Moving Devices with the Cricket Location System Using Polynomial Approximation as Compression and Aggregation Technique in Wireless Sensor Networks Bouabdellah KECHAR bkechar 2000@yahoo. fr Oran University Faculty of science – Department of Computer Science Algeria June 4, 2007 Workshop on Wireless Sensor Networks Marrakech - Morocco

Outlines Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 2

Introduction (1) Characteristics of WSN Important density Limited processing speed Limited storage capabilities Limited power supply (energy) And limited bandwidth Values referenced here are resources available in MICA 2 mote Need design and development of new protocols and algorithms at each level of WSN-layers stack (independently or using Cross layer approach) in order to minimize the dissipated power and consequently extend network lifetime B. KECHAR WWSN – Marrakech – June 4, 2007 3

Introduction (2) The reduction of the volume of data to be transmitted in WSN constitutes the most convenient method to reduce energy consumption in a WSN. This is motivated usually by the fact that processing data consumes much less power than transmitting data. One way to achieve this goal is : Data Compression and Aggregation B. KECHAR WWSN – Marrakech – June 4, 2007 4

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 5

Objective (1) Network stack Collected data (fixed or variable window) Compression & aggregation Transport Layer Sensor stack Multihop Routing Protocol Sensor Layer WSN-MAC Layer Physical Layer Transceiver Unit IN Sensor Channel OUT Polynomial approximation algorithms and Local aggregation Polynomial packet Wireless Channel Temperature, relative humidity, wind speed, … (Environmental readings) B. KECHAR WWSN – Marrakech – June 4, 2007 6

Objective (2) Applications concerned ? Environmental monitoring Temporal constraint is not required Nature of analysis is qualitative Resolution method ? Approach based on theorem of Stone-Weierstrass (theory of approximation of functions) Compression Protocol based on calculation of correlation coefficients between polynomials Local aggregation Validation method ? Simulation using Matlab tool B. KECHAR WWSN – Marrakech – June 4, 2007 7

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 8

Related works LTC (Lightweight Temporal Compression) [Schoellhammer & al 2004] PREMON (PREdiction-based MONitoring) [Goel & al 2001] Ti. NA (Temporal in-Network Aggregation) [Sharaf & al 2003] CAG (Clustered AGgregation) [Sun. Hee & al 2005] TREG (TREe based data a. Ggregation) [Torsha & al 2005] B. KECHAR WWSN – Marrakech – June 4, 2007 9

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 10

Requirements and Constraints Temporal coherency in physical phenomenon Environmental data as temperature, humidity and others, have a common property : continuous variation in time for relatively small temporal windows. The evolution of these properties is roughly linear this characteristic of natural phenomena allows designers of applications to adapt the model of data collection. Interpolation and approximation Stone-Weierstrass theorem Application scenario and suppositions Every sensor have: CPU, RAM, RADIO, protocols Variation of error tolerated by application B. KECHAR WWSN – Marrakech – June 4, 2007 11

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 12

Algorithms of compression : Fixed Window Dim: number of readings by sequence Tab: collected readings Poly: polynomial coefficients ni: sensor node object Var. Err: evaluation of polynomial and calculation of variation m: Polynomial degree Sensed and collected data at time tj B. KECHAR WWSN – Marrakech – June 4, 2007 13

Algorithms of compression : Fixed Window Dim: number of readings by sequence Tab: collected readings Poly: polynomial coefficients ni: sensor node object Var. Err: evaluation of polynomial and calculation of variation m: Polynomial degree Variation of error B. KECHAR WWSN – Marrakech – June 4, 2007 14

Algorithms of compression : Fixed Window Dim: number of readings by sequence Tab: collected readings Poly: polynomial coefficients ni: sensor node object Var. Err: evaluation of polynomial and calculation of variation m: Polynomial degree Find a new polynomial while condition is true, otherwise save polynomial and transmit it B. KECHAR WWSN – Marrakech – June 4, 2007 15

Algorithms of compression : Fixed Window Dim: number of readings by sequence Tab: collected readings Poly: polynomial coefficients ni: sensor node object Var. Err: evaluation of polynomial and calculation of variation m: Polynomial degree Start Approximation using Least-Squares method B. KECHAR WWSN – Marrakech – June 4, 2007 16

Algorithms of compression : Variable Window Dim: number of readings by sequence Tab: collected readings Poly: polynomial coefficients ni: sensor node object Var. Err: evaluation of polynomial and calculation of variation Window. Min: minimal size of time window Window. Max: maximal size of time window Old. Degree: Degree of last approximation m: Polynomial degree B. KECHAR WWSN – Marrakech – June 4, 2007 17

Algorithms of compression : Variable Window Sensed and collected data of initial window B. KECHAR WWSN – Marrakech – June 4, 2007 18

Algorithms of compression : Variable Window Initialization B. KECHAR WWSN – Marrakech – June 4, 2007 19

Algorithms of compression : Variable Window Check if the old polynomial is extensible B. KECHAR WWSN – Marrakech – June 4, 2007 20

Algorithms of compression : Variable Window A new collected value is added and the old degree is saved B. KECHAR WWSN – Marrakech – June 4, 2007 21

Algorithms of compression : Variable Window To limit the algorithm by a number of readings (Window. Max) B. KECHAR WWSN – Marrakech – June 4, 2007 22

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 23

Local Aggregation Coefficient of correlation With IDS tn V 1 …. . Vn Packet structure Without compression tn P(ti) Packet structure With compression Correlated polynomial Transmit juste B. KECHAR IDS WWSN – Marrakech – June 4, 2007 IDS tn Packet structure With compression 24

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 25

Experiments and Simulation (1) Compression ratio during one period Algorithm with Fixed Window Experiment: samples of 1000 readings (experimental, Temperature, Humidity and Wind speed) Environmental Real values If we increase the number of readings, that do not imply automatically a corresponding better rate. Contrary, when the window sizes are reduced, the correlation is very expressive and then the approximation process is better. Compression Quality vs Window Size With Tolerable error variation =0. 1 B. KECHAR WWSN – Marrakech – June 4, 2007 26

Experiments and Simulation (2) Algorithm with Variable Window Compression ratio fully depends on the tolerable variation of error, which implies the strong connection between the quality of data and the desirable compression ratio. Compression Quality vs Tolerable variation of error This table shows that the majority of the values reconstituted by the evaluation of the polynomials will be in the specified margin Temperature Tolerable Error Variation Humidity Wind Speed 0. 1 Compression Rate 15. 09% 75. 92% 64. 21% Restitution Rate 99. 98% 100% 99. 80% Restitution Rate B. KECHAR WWSN – Marrakech – June 4, 2007 27

Experiments and Simulation (3) Comparison of compression rate ASFW Algorithm ASVW Algorithm Experimental Data Temperature Humidity Wind Speed 78 % 27. 55 % 83. 37 % 80. 83 % 15. 14 % 15. 09 % 75. 92 % 64. 21 % If we fix the variation of error at 0. 1 and we consider an optimal size of the fixed window (80 readings) for the algorithm with fixed window, the algorithm with variable window is more powerful. B. KECHAR WWSN – Marrakech – June 4, 2007 28

Contents Introduction Objective Related works Requirements and constraints Algorithms of compression Fixed window based Variable window based Local aggregation Experiments and simulation Conclusion and future works B. KECHAR WWSN – Marrakech – June 4, 2007 29

Conclusion Data compression is an important technique to reduce communications and hence save energy in WSN. Our proposed approach (New data compression and aggregation technique for WSN) is a simple idea but it is quite novel and interesting. The results obtained are encouraged to follow this research direction. B. KECHAR WWSN – Marrakech – June 4, 2007 30

Perspectives What are the computation cost and memory requirement at each sensor node ? A comparison with other compression techniques in terms of accuracy and cost (like Ti. NA and LTC). Additional experimental effort to prove the effectiveness of the approach (Energy calculation). Extend the approach to Multi-objective WSN (several data types in the same network with cooperation capabilities) B. KECHAR WWSN – Marrakech – June 4, 2007 31

Using Polynomial Approximation as Compression and Aggregation Technique in Wireless Sensor Networks Thanks for your attention. Questions & remarks B. KECHAR WWSN – Marrakech – June 4, 2007 32