Aumento de confiabilidade de sistemas embutidos usando redundância e algoritmos de decisões baseados em reconhecimento de padrões

Abstract

The continuous progress in integrated circuit technologies leads to a constant increase of the number of embedded systems surrounding us. Amongst the several challenges accompanying this trend, system robustness has an outstanding role. This is even more true for critical applications that are being controlled by embedded systems. Most of these systems interact with the world by means of information coming from data acquisition from sensors. Consequently, the correct system behavior depends on a reliable data acquisition. In case of faulty data, it is mandatory to possess solutions that prevent that these faults harm the system or even generate disasters. This dissertation presents an approach that improves the reliability of embedded systems that apply sensor data. The proposed technique uses a mixture of redundancy techniques and detection of abnormalities based on Pattern Recognition. In order to verify the proposed approach, a heating pilot plant has been developed. The experiments were performed with online data for redundancy techniques and offline for pattern recognition techniques. The results indicate the feasibility of the proposed methods.