Structures for In-Network Moving Object Tracking in Wireless Sensor Networks Chih-Yu Lin and Yu-Chee Tseng Department of Computer Science and Information Engineering National Chiao Tung University BROADNETS 2004 Speaker: Hsu-Ruey Chang
Outline Introduction Problem statement Tree Construction Algorithms A Greedy Deviation Avoidance Tree A Zone-based Deviation-Avoidance Tree Simulation Results Conclusions
Introduction Sensor Network Computing power Storage space One important application of wireless sensor networks is tracking moving objects Location update Location query
Introduction Object tracking Localized prediction approach Cooperative tracking Tree architecture Convey tree Message pruning tree DAT (Deviation-Avoidance Tree) Z-DAT (Zone-based DAT)
Problem Statement Goal Not to propose a location-tracking model Proposed a data-aggregation model for this kind of service
Problem Statement VGVG EGEG W G (A,B)
Problem Statement Our goal is to construct from G a logical weighted tree Message-pruning tree The total communication cost is as low as possible
Problem Statement VTVT ETET W T (A,B)
Problem Statement A cost function of T by counting the number of events transmitted in the network
Problem Statement
A Greedy Deviation-Avoidance Tree Observation 1 From Eq. 1, we observe that the minimal value of dist T (u, par(u, v)) is dist G (u, par(u, v)) We say that T is deviation-free Fig. 4(a), (c), and (d)
A Greedy Deviation-Avoidance Tree Observation 2: From Eq. 2, we observe that the minimal value of w T (u, v) is 1 when u ≠ v, i.e. not only (u, v) ∈ E T but also (u, v) ∈ E G Therefore, we would expect that each sensor’s parent is its neighbor Fig. 4(d) Conducting this observation to Eq. 1, it means that the average value of dist T (u, par(u, v))+dist T (v, par(u, v)) is reduced
A Greedy Deviation-Avoidance Tree Observation 3: w G (u, v) > w G (u, v) We would expect that dist T (u, par(u, v)) + dist T (v, par(u, v)) < dist T (u, par(u, v)) + dist T (v, par(u, v)) Based on this observation and the second observation, an edge (u, v) with a higher w G (u, v) should be merged into T early and par(u, v) should be either u or v
A Greedy Deviation-Avoidance Tree
A Zone-based Deviation-Avoidance Tree Consider the term The perimeter that bounds the area covered by sensors in Subtree(v) may have a significant impact on the cost function C(T)
A Zone-based Deviation-Avoidance Tree
Simulation Results In the first model We deploy 4096 sensors in a 64 × 64 grid network, one in each grid In the second model We consider a 256 × 256 grid network in which 4096 sensors are randomly deployed
Simulation Results We consider two performance metrics Update cost C(T) Querying cost Q(T)
Simulation Results
Conclusion We have presented two message-pruning structures for moving object tracking in a sensor network We are currently investigating more mobility models other than the city mobility model to observe their effects