Rapidly-Exploring Tree With Linear Reduction: A Near-Optimal Approach for Spatiotemporal Sensor Deployment in Aquatic Fields Using Minimal Sensor Nodes

Spatiotemporal monitoring of large fields for water quality motivates this paper. The primary goal of this paper is to find a suitable deployment strategy for mobile sensor nodes in aquatic fields, which receives the least estimation error with minimal sensor nodes. Typically, the resources of mobile sensor nodes are limited in the case of large field monitoring, so they cannot be deployed arbitrarily. Hence, an optimal sensor node deployment strategy with minimally required sensor nodes becomes a primary focus of this paper. To address this problem, an optimal sensor node deployment strategy termed Rapid Random exploring tree with Linear Reduction (RRLR) is developed in this paper. It relocates sensor nodes to their best sensing locations while reducing the required number of sensor nodes without losing information. The approach is based on an environmental model and the linear dependence of sensor readings. In particular, the spatiotemporal correlation of the sensor node deployment in a large geographic area of interest is minimized. It is also proved that the optimization objective function, which uses the prior estimation error, is submodular and results in a near-optimal solution. Simulation results show that RRLR requires much fewer sensor nodes to achieve the same or lower estimation error when compared to benchmark algorithms.