Energy-Aware Scheduling with Quality of Surveillance Guarantee in Wireless Sensor Networks Jaehoon Jeong, Sarah Sharafkandi and David H.C. Du Dept. of Computer Science and Engineering, Univ. of Minnesota International Conference on Mobile Computing and Networking Dependability issues in wireless ad hoc networks and sensor networks,
Outline 2 Introduction Related Work Problem Formulation Energy-Aware Sensor Scheduling Optimality of Sensor Scheduling QoSv-Guaranteed Sensor Scheduling Sensor Scheduling for Complex Roads Performance Evaluation Conclusion
Introduction 3 Motivation We investigate the properties of the Linear Sensor Network (e.g., Road Network in transportation system). These properties can be used for a variety of applications: Localization, Vehicle Detection, and Vehicle Tracking. Applications of This Sensing Scheduling Algorithm Surveillance for Security around City’s Border Crossroad Signal Control in Transportation System Objectives Maximization of Lifetime of Wireless Sensor Network Control of Detection Quality Quality of Surveillance Guarantee (QoSv) Contributions Energy-aware Sensor Scheduling feasible for Mobile Target Detection and Tracking QoSv-Guaranteed Sensor Scheduling for Complex Roads
Surveillance of City Border Roads (1) 4
Surveillance of City Border Roads (2) 5
Vehicle Detection for Crossroad Signal Control 6
Related Work 7 Temporally and Spatially Partial Coverage The region under surveillance is covered partially in terms of time and space. Our scheduling algorithm utilizes this partial coverage to save sensing energy. Quality of Surveillance (QoSv) Our QoSv is defined as the reciprocal of the average detection time. Other QoSv was originally defined as the reciprocal value of the expected travel distance until the first detection.
Problem Formulation 8 Assumptions The sensors knows their location and are time-synchronized. The sensing range is uniform-disk whose radius is r ( r is longer than a half of the road’s width). The cost of turn-off operation is ignorable. The vehicle’s maximum speed is bounded as: Objective To maximize the sensor network lifetime to satisfy the following conditions Provide the reliable detection of every vehicle Guarantee the desired average detection time Facilitate the mobile target tracking after the target detection.
Sensor Network Model for Road Segment 9
Key Idea to This Scheduling 10 How to have some sleeping time to save energy? We observe that the vehicle needs time l/v to pass the road segment. Time l/v is the sleeping time for all the sensors on the road segment.
Energy-Aware Sensor Scheduling 11 Our sensor scheduling consists of two phases: Initialization Phase Surveillance Phase Working Period + Sleeping Period
Sensing Sequence for Vehicle Detection 12
Optimality of Sensor Scheduling (1) 13 Sensor Network Lifetime The following energy can be saved through sleeping: Number of Schduling Periods Working Period Sleeping Period n : total number of sensors w : working time of sensor l : length of road v : max possible vehicle speed : lifetime of each sensor
Optimality of Sensor Scheduling (2) 14 Schedule1 is this outward unidirectional scheduling, and Schedule2 is an optimal scheduling Inequality of lifetime which results in Actually, X should be equal to the number of working periods because after each sleeping period there should be a working period Schedule1 is optimal scheduling X : number of sleeping periods l/v : upper bound on the sleeping period
Considerations on Turn-On and Warming-UP Overheads 15 Each Sensor’s Lifetime without Sleeping Sensor Network Lifetime through Sleeping Case 1: Turn-On Overhead is greater than Sleeping benefit Case 2: Turn-On Overhead is less than Sleeping benefit t : min time needed for each sensor to detect and transmit data
QoSv-Guaranteed Sensor Scheduling 16 Average Detection Time for Constant Vehicle Speed Approximate Average Detection Time (ADT) Average Detection Time for Bounded Vehicle Speed
Determination of Scheduling Parameters 17 Scheduling (under sensing error) Parameters are Sensor Network Length ( l ) Working Time ( w ) Sleeping Time ( s ) where m : the number of scanning per working period P success : the success probability of one scanning
Sensor Scheduling for Complex Roads (1) 18 Road Network between the Inner and Outer Boundaries
Sensor Scheduling for Complex Roads (2) 19 A Connected Graph for an Exemplary Road Network The Road Network is represented as a Connected Graph between the Inner and Outer Boundaries.
Sensor Scheduling for Complex Roads (3) 20 Construction of Scheduling Plan in Road Network Determine the starting points Si to satisfy the required QoSv through Search Algorithm.
Sensor Scheduling for Complex Roads (4) 21 Scanning in Road Network One scanning can be split into multiple scanning. Multiple scanning can be merged into one scanning for sensing energy.
Performance Evaluation 22 Metrics Sensor Network Lifetime according to Working Time and Turn-on Energy Average Detection Time according to Working Time and Road Segment Length (i.e., Sensor Network Length) Required Average Scanning Number for Sensing Error Probability Validation of Numerical Analysis We validated our numerical analysis of our scheduling algorithm through simulation.
Environment for numerical analysis 23 road segment’s width is 20m, length is 2000m number of sensors is 100 total sensing energy in each sensor is 3600J, can used continuously for 3600sec since sensing energy consumption rate is 1watts working time per working period is in [0.1, 5] turn-on energy consumption is {0,0.12,0.48,0.96}J vehicle’s max speed is 150km/h
Sensor Network Lifetime according to Working Time and Turn-on Energy 24
Average Detection Time according to Working Time and Road Segment Length 25
Required Average Scanning Number for Sensing Error Probability 26
Conclusion 27 We proposed an Energy-Aware Scheduling Algorithm to satisfy the required QoSv in Linear Sensor Network. QoSv is defined as the reciprocal value of Average Detection Time (ADT). Our Algorithm can be used for Surveillance for City’s Border Roads, and Traffic Signal Control in Crossroads Future Work Enhance the scheduling scheme when the sensors are deployed randomly close to the roads Extend the scheme to two-dimensional open field