Cognitive quality of service predictions in multi-node wireless sensor networks

Wireless sensor networks lead the way to the realization and penetration of the internet of things into daily life. As the sensors and communication devices interconnect things and people more, the quality of service demands become stringent and diverse. Optimizing conflicting quality of service goals is NP-hard. Moreover, the need for the communication systems to dynamically adapt has grown as factors like scale, application-specific performance demands, and deployment scenarios evolve. In this article, we intend to predict the quality of service with an aim to improve the performance of services in the internet of things. We conducted experiments using a real test-bed and collected performance data under a wide range of communication parameter configurations. Statistical analysis revealed a significant relationship between communication parameters and quality of service metrics. Based on the correlations, we trained deep learning models to assess the predictability of the metrics. The prediction results are encouraging. The outcomes of this research pave the way to lay the foundation for a data-driven design of adaptive quality of service in wireless sensor networks and the internet of things.