80 likes | 204 Vues
This document explores the efficient management and analysis of moving objects through a specialized tree structure. It covers vital aspects like query types and indexing, object velocity, and lifespan, while ensuring a compact tree size. Key focus areas include handling various queries such as retrieving future intersection data within specified timeframes and the need for both predictive and historical information. The framework leverages modifications of existing TPR* and HR+ tree algorithms for operations like insert, update, and delete, ensuring optimized performance in dynamic environments.
E N D
HTPR* TreeRequirement Analysis CSE 19981234 김승준
Contents • Moving Objects • Query Type and Indexing • Velocity • Lifespan • Tree Structure • Insert, delete, update
t t y y x x Moving Objects • Moving points • Change the position of object by time • Moving regions • Change the position and shape of object by time
Query Type and Indexing • Past, Current and Future query • Future : “Retrieve the objects that will intersect the polygon P within the next 3 minutes” • Indexing • Need predictive information • Need historical information • Goal • Cover all query type with single tree structure • Tree structure should be not too large size
Velocity Vector • Object with linear motion • Velocity vector can change!
Lifespan • When velocity changed • Lifespan 0 to 1, node ‘b’ have V(1,1,1,1) • Lifespan 1 to *, node ‘b’ have V(0,-2,0,-2)
Tree Structure • Root node • (I, V, t, p) • I : MBR (Minimum Bounding Rectangle) • V : Velocity Vector • t : Timestamp • p : pointers to child node • Other node • (I, V, tstart, tend, p) • tstart : timestamp start modifier • tend : timestamp end modifier
Insert, Update, Delete • Insert • Insertion is similar to TPR* tree except consideration about lifespan • Update • When velocity changed, lifespan of old node is updated and duplicated node with new lifespan and velocity is inserted • Delete • Deletion can only happens when lifespan start & end time is same • Ex) (tstart = 5, tend = 5) • All action is based onto TPR* & HR+ tree algorithm and needs somewhat modification