Utilizing Correlations to Compress Time-Series in Traffic Monitoring Sensor Networks

Abstract

Wireless sensor networks present significant opportunities for fine-grained and continuous monitoring of road traffic, enabling careful city planning, automated road maintenance and accident detection. Users are typically willing to tolerate a small error in car-flow data, in order to reduce the cost of data propagation from the sensor nodes to the gateway nodes, to which users are connected. In this paper, we first examine the relative performance of Fourier- and wavelet-based algorithms for compressing traffic data locally at the sensor nodes. Using real traffic information from the city of Cambridge (UK), we then demonstrate that car-flow data collected across geographically dispersed sensor nodes exhibit strong spatial and temporal correlations. We then combine lossy Fourier-compression with correlation-based compression to achieve further communication savings within a user-specified error threshold. For a tolerated error of 5-15 cars per 5 min, it is shown that exploitation of temporal correlations yields 14-30% savings relative to Fourier compression alone, whilst use of spatial correlations results in 10-35% savings.