Self-repair of embedded systems

As embedded systems are increasingly entrusted with critical tasks, it becomes important that these systems exhibit high levels of reliability. Increasing complexity makes production of fault-free systems extremely difficult. Developing self-repair for embedded systems provides a means to alleviate the effects of a fault. This paper discusses a self-repair methodology for embedded systems that employs the unified modelling language and genetic algorithms. This approach increases the flexibility of systems by enabling the transfer of functionality between hardware and software. This design methodology is demonstrated using a speech recognition system. The benefits of the design for a self-repair strategy are highlighted and the implementation results for two examples of design objectives are presented. [Copyright &y& Elsevier]