Abordagens para reconfiguração de sistemas de tempo real com QoS e restrições de energia e temperatura


This thesis proposes a framework for dynamic reconfiguration, value-based processor resource allocation in multi-modal or not real-time applications, under schedulability, energy consumption and temperature constraints. The framework is suitable for critical and soft real-time adaptive embedded systems which need guarantees of energy savings. The dynamic allocation is formulated as a discrete and continuous (convex and piecewise linear) optimization problem for which efficients algorithms were tested. Although the discrete problem is NP-Hard, the others have efficient solution and numerical analysis and simulations have shown that the used algorithms and models achieves very good results, with low computational cost.

Download (Brazilian Portuguese)