Soft Network Organisation Towards Future Distributed ML-based Sensing Systems
Krzysztof Cichoń · 2020
The article analyses efficient information detection by a network of sensors. It is assumed that every sensor collects local information, however, with the application of clustering in the network, it is possible to efficiently exchange knowledge within the network. In particular, the local knowledge about a licensed signal is acquired by sensors, and it is then exchanged and transferred to global knowledge. To this end, two types of clustering are analysed: hard K-means clustering and soft Kmeans. In the latter case, the crucial knowledge about stiffness is important to perform an efficient exchange of information. The analysis of desired stiffness values has been carried out with simulation experiments where the energy usage and reporting efficiency were compared for various sizes and types of clusters. Moreover, the quality of global knowledge is also shown in the simulation results.