Contributions to industrial wireless sensor networks.

Abstract

Recent advances in wireless sensor networks (WSN), especially in industrial environments (IWSN), have brought important improvements for the deployment of a sensor network in such environments. However, there is a high level of interference, noise, shadowing, and multipath fading in industry, caused by machinery, metallic objects and obstructions, which can affect the quality of the link. Some mechanisms have been proposed to mitigate the negative effects, such as frequency hopping and channel blacklisting. However, there are still open issues, such as an appropriate method of managing both mechanisms to avoid similar or adjacent channels in simultaneous transmissions. The use of a link quality estimator (LQE), with adequate channel assignment, helps to increase network performance. In addition to the use of estimators, this thesis proposes two approaches that separate the channels according to the temporary profile of each one and insert them into lists. The aim is to better deal with the negative effects of the environment on the transmission medium. The first uses a double channel list (alowlist and denylist), and the second uses a triple list of channels (denylist, greylist, and alowlist). In both, the management of channel offset was considered to avoid collisions and reduce internal interference in simultaneous transmissions. The denylist has low quality channels, the greylist has channels with uncertain quality, and the alowlist has good quality channels. A fuzzy logic method was used in the triple list approach to classify the channels in the most appropriate list. The two proposals were compared by means of simulation using the frequency hopping method based on the TSCH protocol of the IEEE 802.15.4e standard for WSN. In the study, networks in star and tree topologies were analyzed, with and without link quality estimators, with and without collisions, using a realistic channel model for IWSN. The results of the experiments showed that in the tree topology the approaches with double and triple list presented better performances than the approach without link quality estimators, in terms of package delivery rate and determinism, and that the triple list surpassed the double list in more dynamic networks. In star topology, the larger the size of the denylist, the better the network performance, as the devices use only the best channels.