Robust data collection for energy-harvesting wireless sensor networks

Energy-harvesting wireless sensor networks (EHWSN) have drawn much attention in recent years because the capability of collecting ambient energy enables the perpetual operations of sensor nodes. However, the instability of renewable energy sources has also imposed new challenges to data collection in EHWSNs. In order to achieve perpetual operation, many studies have proposed adjusting the sensors sampling rates or reconfiguring the underlying routing structure to counter the effects of these challenges. However, the performance of the former is constrained and sensitive to the routing structure used, while the latter requires global signaling, which can interrupt network operations. In this paper, we propose to address the dynamics of renewable energy with a two-stage approach. In the network planning stage, we make use of the primal cut method to solve a two-stage robust optimization (RO) problem and construct a data collection tree that works well under all worst-case scenarios. While in the operational stage of the network, we propose another algorithm that can lexicographically maximize the sampling rates of sensor nodes according to the observed recharging rates with minimal overheads. This avoids reconfiguring the routing structure during the operational phase of the network while simultaneously maximizes the performance of the network under the uncertainty of renewable energy. Numerical results are presented to show the effectiveness and robustness of the proposed method in dealing with the variability of renewable energy.