| บทคัดย่อ(English) |
This project presents Genetic Algorithm (GA) technique with a dynamic penalty function for solving a reliability optimization problem of embedded (hardware and software) systems, using redundancy techniques, which are N-Version Programming and Recovery Blocks, considering system cost constraints. Our approach is the first time that GA is applied to optimize this type of systems, where related faults or dependency in software components / versions are considered. It is our extended work from N. Wattanapongsakorn and Levitan [9] where Simulated Annealing algorithm was applied to the problem with no guarantee for optimal solutions. GA is a suitable optimization approach since it provides robustness and efficiency in solving combinatorial optimization problems with large, and complex search spaces. The solutions are compared with a non-linear programming technique, where optimal solution can be obtained and identified. This project consists of two main parts. In the first part, we consider reliability optimization problems assuming that hardware and software component reliability are constant and known. In the second part, we extend our work by considering the same problems with component reliability estimation uncertainty where mean and its variance are captured. |
