This article describes how fault tolerance is an essential issue for many WSN (Wireless Sensor Network) applications such as wildlife monitoring, battlefield surveillance and health monitoring. It represents a great challenge for researchers regarding to the characteristics of sensor nodes which are prone to failures due essentially to their limited resources. Faults may occur, not only when sensor nodes exhaust their energy, but also when the congestion phenomenon emerges, because of a high traffic in the network and limited storage capacity of the sensor nodes. In order to support fault tolerance in WSNs, the authors propose a new scheme which incorporates a link quality estimation algorithm and a congestion detection mechanism to enable nodes that present high quality links to be chosen for routing in a non-congested area in case of faults. Evaluations through simulations under NS2 show that our proposed protocol tolerates faults with a minimum cost relatively to HEEP protocol and improves network’s performances comparatively to other fault tolerant protocols such as EF-LEACH.

The selectivity is one of the main challenges to develop a gas sensor, the good chemical species detection in a gaseous mixture decreasing the missed detections. The present paper proposes a new solution for gas sensor selectivity based on artificial neural networks (ANNs) and fuzzy logic (FL) algorithm. We first use ANNs to develop a gas sensor model in order to accurately express its behavior. In a second step, the FL and Matlab environment are used to create a database for a selective model, where the response of this one only depends on one chemical species. Analytical models for the gas sensor and its selective model are implemented into a Performance Simulation Program with Integrated Circuit Emphasis (PSPICE) simulator as an electrical circuit in order to prove the similarity of the analytical model output with that of the MQ-9 gas sensor where the output of the selective model only depends on one gas. Our results indicate the capability of the ANN-FL hybrid modeling for an accurate sensing analysis.