An Agent Based Data Collection Scheme for Vehicular Sensor Networks

Abstract

A Vehicular Sensor Network (VSN) is a sensing platform composed of smart onboard sensor nodes (vehicles) and roadside units, in which vehicles continuously collect sensor data from the road network to enable a range of real-time data-intensive applications, such as traffic pattern/congestion analysis, road surface diagnosis, and urban pollution monitoring. However, due to lossy links, limited bandwidth and highly dynamic network topology, it is very challenging to efficiently collect the data generated by vehicles on the road, especially under dense traffic situations. In this paper, we propose to deploy mobile agents for collecting sensor readings from a given road segment of interest. The mobile agent migrates among vehicles within the segment via wireless broadcast and uses local on-board computational resources to process and collect data as required. Since the wireless links are generally lossy, a broadcast may not reach all the vehicles within the segment; thus, to improve the reliability of the scheme, we further propose a termination decision algorithm based on recursive Bayesian estimation by which the agent decides whether all vehicles within the segment have been visited. Extensive simulation results show that the proposed agent-based data collection scheme achieves close to 100% data collection coverage under a wide range of vehicular traffic densities, while retaining a small communication overhead.