Efficient computation of wireless sensor network lifetime through deep neural networks

The most important quality-of-service metric for wireless sensor networks (WSNs), arguably, is the lifetime. Estimating the network lifetime under optimal operation conditions can be done by casting the problem as a mixed integer programming (MIP) model and solving the problem instances to optimality. Yet, solution times of such models are excessively high. Therefore, it is not possible to work with large problem instances within an acceptable solution time. Adopting learning based algorithms has the ability to produce near-optimal results much more rapidly in comparison to MIP models. In this study, we propose a deep neural network (DNN) based model to determine the WSN lifetime near-optimally virtually instantly. The proposed model is able to predict the lifetime of a randomly deployed WSN over a predetermined area with an average accuracy more than 98.5%. An interesting outcome of the study is that the DNN based model is able to estimate the lifetime of WSNs with higher number of nodes successfully even if it is trained with a dataset obtained with lower number of nodes.

Automated summary

This study proposes a deep neural network-based model to determine the lifetime of wireless sensor networks near-optimally almost instantly, with an average accuracy exceeding 98.5%. Learning-based algorithms are able to produce near-optimal results more rapidly compared to mixed integer programming models.