Evolutionary Fault Repair of Electronics in Space Applications

The project involved familiarizing with the field of artificial evolution, in particular its applications in the space industry. A field was found where it would be possible to add to general knowledge by conducting experiments designed by the student, namely the issue of fault tolerance in space applications, with particular emphasis on the use of evolutionary techniques. The experiments involved enhancing a much used method for fault tolerance by adding a genetic algorithm. Here, I present only my final ideas, experiments and conclusions, and not everything along the winding path to get there. For example, several ideas for experiments on similar topics were developed and discarded before the project finally took its current form. I would like to thank my supervisors Adrian Thompson and Pauline Haddow for their support in my work. Abstract This project has explored the use of genetic algorithms (GAs) in the field of fault tolerance in electronics. Special emphasis has been put on fault tolerance in space applications, a field where autonomous systems are important and faults can have great implications. It is the author's view that this field can benefit greatly from an evolutionary approach. Several experiments were conducted to explore the possibility of using evolution for fault repair. These experiments indicate that repair of digital logic is a hard task for general genetic algorithms. An enhanced version of a fault tolerance scheme commonly used in space applications (a voting system) was proposed. The experiments show that although genetical fault repair failed to impress, the newly proposed system performs potentially very well. The report also highlights some practical issues that should be investigated before the idea will be applicable in real systems.