Evaluation of the Use of Compressed Sensing in Data Harvesting for Vehicular Sensor Networks

Abstract

We propose a new harvesting approach for Vehicular Sensor Networks based on compressed sensing (CS) technology called Compressed Sensing-based V ehicular Data Harvesting (CS-VDH). This compression technology allows for the reduction of the information volume that nodes must send back to the fusion center and also an accurate recovery of the original data, even in absence of several original measurements. Our proposed method, thanks to a proper design of a delay function, orders the transmission of these measurements, being the nodes farther from the fusion center, the ones starting this transmission. This way, intermediate nodes are more likely to introduce their measurements in a packet traversing the network and to apply the CS technology. This way the contribution is twofold, adding different measurements to traversing packets, we reduce the total overload of the network, and also reducing the size of the packets thanks to the applied compression technology. We evaluate our solution by using ns-2 simulations in a realistic vehicular environment generated by SUMO, a well-known traffic simulator tool in the Vehicular Network domain. Our simulations show that CS-VDH outperforms Delay-Bounded Vehicular Data Gathering (DB-VDG), a well-known protocol for data gathering in vehicular sensor networks which considers a specific delay bound. We also evaluated the proper design of our delay function, as well as the accuracy in the reconstruction of the original data. Regarding this latter topic, our experiments proved that our proposed solution can recover sampled data with little error while still reducing the amount of information traveling through the vehicular network.