Environmental Monitoring via Vehicular Crowdsensing

Spatial field reconstruction is important for a variety of applications based on environmental monitoring via Internet-of-Things and vehicular communications. This work develops a framework for the analysis of multidimensional stochastic sampling in vehicular crowd sensing, where samples are gathered from sensors on vehicles. Vehicular crowd sensing performance in terms of reconstruction mean-square error is compared to that obtainable with fixed installation. In the first case, sensors are assumed to be randomly distributed over the monitored area, while, in the latter, they are considered as regularly placed in a lattice. In addition, the positions of the mobile nodes are assumed non perfectly known at the interpolator. By extending recent results on multidimensional stochastic sampling, it is shown that a high field reconstruction accuracy can be obtained by vehicular crowd sensing even in cases where fixed infrastructure would lead to insufficient sampling.