N-Version Design Versus One Good Version
IEEE Software
WORDS '01 Proceedings of the Sixth International Workshop on Object-Oriented Real-Time Dependable Systems (WORDS'01)
Diagnosis and Fault-Tolerant Control
Diagnosis and Fault-Tolerant Control
Explorations in design space: unconventional electronics designthrough artificial evolution
IEEE Transactions on Evolutionary Computation
| Hi-index | 0.00 |
The multi-version approach to the system design is becoming a standard for the critical applications as it allows decreasing the probability of common mode failure. The main idea of this approach is to obtain the least correlated versions of the same project. The effective way to satisfy this requirement is to combine classical and non-classical approaches to the system design. In this point, genetic algorithms (GAs) can be successfully used as one of the non-classical approaches in order to evolve the non-conventional digital systems. However, there are two main issues that arise while using GA: in some cases GA is not able to evolve a completely correct solution; if a complexity of system logic grows the time consumptions significantly increase. To avoid both issues, we elaborate the concepts of correctness and definiteness for digital automata, which are the solutions received with GA, and represent the efficient design method based on the GA application. It allows obtaining fault-tolerant and compact digital systems with partially correct and partially definite automata. The new approach has been illustrated by the example of the heating controller for AN-70 plane.