Efficient Threshold Aggregation of Moving Objects

TL;DR

This paper introduces efficient threshold-based aggregation operators for moving objects, enabling fast estimation of congestion metrics in dynamic query spaces with high accuracy, applicable to ecological and safety domains.

Contribution

The paper presents novel threshold aggregation operators for moving objects, transforming position-based selection into threshold-based selection, with algorithms and experimental validation.

Findings

Operators produce results with under 5% error.

Algorithms are efficient for large query sets.

Applicable to ecological monitoring and aviation safety.

Abstract

Calculating aggregation operators of moving point objects, using time as a continuous variable, presents unique problems when querying for congestion in a moving and changing (or dynamic) query space. We present a set of congestion query operators, based on a threshold value, that estimate the following 5 aggregation operations in d-dimensions. 1) We call the count of point objects that intersect the dynamic query space during the query time interval, the CountRange. 2) We call the Maximum (or Minimum) congestion in the dynamic query space at any time during the query time interval, the MaxCount (or MinCount). 3) We call the sum of time that the dynamic query space is congested, the ThresholdSum. 4) We call the number of times that the dynamic query space is congested, the ThresholdCount. And 5) we call the average length of time of all the time intervals when the dynamic query space is congested, the ThresholdAverage. These operators rely on a novel approach to transforming the problem of selection based on position to a problem of selection based on a threshold. These operators can be used to predict concentrations of migrating birds that may carry disease such as Bird Flu and hence the information may be used to predict high risk areas. On a smaller scale, those operators are also applicable to maintaining safety in airplane operations. We present the theory of our estimation operators and provide algorithms for exact operators. The implementations of those operators, and experiments, which include data from more than 7500 queries, indicate that our estimation operators produce fast, efficient results with error under 5%.