170 likes | 293 Views
Optimal Tracking Interval for Predictive Tracking in Wireless Sensor Network. IEEE COMMUNICATIONS LETTERS, VOL. 9, NO. 9, SEPTEMBER 2005 Zhen Guo, Mengchu Zhou, Fellow, IEEE, ( 周孟初 , http://web.njit.edu/~zhou/) and Lev Zakrevski, Member, IEEE Presentation by Cheng-Ta Lee. Outline.
E N D
Optimal Tracking Interval forPredictive Tracking in Wireless Sensor Network IEEE COMMUNICATIONS LETTERS, VOL. 9, NO. 9, SEPTEMBER 2005 Zhen Guo, Mengchu Zhou, Fellow, IEEE, (周孟初, http://web.njit.edu/~zhou/) and Lev Zakrevski, Member, IEEE Presentation by Cheng-Ta Lee
Outline • Introduction • Predictive Tracking Sensor Network Architecture • Power Optimization and Quantitative Analysis • Conclusion • Future Work
Introduction 1/3 • Object tracking is an important application in wireless sensor networks • Terrorist attack detection • Traffic monitoring • Most of researchers concentrate on tracking objects and findingefficient ways to forward the data reports to the sinks
Introduction 2/3 • Tracking Interval • As the tracking interval becomes lower↓, in other words ”more frequent↑”, the tracking power consumption is increased ↑ • As it increases ↑, the miss probability increases ↑, thereby lowering the tracking quality ↓
Introduction 3/3 • This paper intends to • propose a quantitative analytical model to find such an optimal tracking interval • study the effect of the tracking interval on the miss probability • propose a scheme called Predictive Accuracy-based Tracking Energy Saving (PATES) by exploiting the tradeoff between the accuracy and cost of sensing operation.
Predictive Tracking Sensor Network Architecture 1/2 • Object Tracking Sensor Networks • An object tracking sensor network refers to a wireless sensor network designed to monitor and track the mobile targets in the covered area • Generally, each sensor consists of three functional units • Micro-Controller Unit (MCU) • Sensor component • RF radio communication component
Predictive Tracking Sensor Network Architecture 2/2 • Predictive Accuracy-based Tracking Energy Saving (PATES) • In PATES, three modules must be in use. • Monitoring and tracking • Prediction and reporting • Recovery • The targets are missed, then the recovery module is initiated • ALL NBR recovery • ALL NODE recovery
Power Optimization and Quantitative Analysis 1/6 • quadratic function • s: tracking interval • a, b, and c are the constants • missing probability P(s)
Power Optimization and Quantitative Analysis 2/6 • m: number of the neighbor around the current node. • N: total number of sensors in whole network • Notification: when a neighbor nodes detects the target, it sends notification to the currect node
Power Optimization and Quantitative Analysis 3/6 • T: Entire period • s: Tracking interval
Power Optimization and Quantitative Analysis 5/6 a=0.0013, b=0.025, and c=0.062
Fig. 2 shows the relationship between the power consumption and tracking interval Power Optimization and Quantitative Analysis 6/6
Conclusion • The power consumption with respect to tracking intervals can be minimized with a quadratic miss probability function under a given prediction algorithm • A predictive tracking scheme to optimize the power efficiency with two stages of recovery is proposed • The proposed scheme is demonstrated by the analytical results to be capable of successfully balancing the tradeoff between the prediction accuracy and tracking cost
Future Work 1/2 • Propose an algorithm to automatically model and validate the real-time relationship between miss probability and tracking interval • Consideration three stages recovery or other recovery mechanism (for example, wake up all the two steps’ neighbor nodes around the current sensor in ALL_NBR recovery stage)
missing probability Power consumption Number of wake up the neighbor nodes around the current sensor in next state Number of wake up the neighbor nodes around the current sensor in next state Future Work 2/2 • Decrease missing probability • Because Erecovery= 9656mJ >>Esuccess= 42mJ • For example, (always) wake up all the neighbor nodes around the current sensor in next state (Optimal number of wake up the neighbor nodes around the current sensor in next state)