An Evolutionary Method for Improving the Reliability of Safetycritical Robots against Soft Errors

Abstract

Nowadays, Robots account for most part of our lives in such a way that it is impossible for us<br />to do many of affairs without them. Increasingly, the application of robots is developing fast<br />and their functions become more sensitive and complex. One of the important requirements of<br />Robot use is a reliable software operation. For enhancement of reliability, it is a necessity to<br />design the fault tolerance system. In this paper, we will present a genetic algorithm and<br />learning automata with high reliability to evaluate the software designed into the robot<br />against soft-error with minimum performance over-head. This method relies on experiment;<br />hence, we use the program sets as criteria in evaluation stages. Indeed, we have used the error<br />injection method in the execution of experimental processes. Relevant data, regarding<br />program execution behavior were collected and then analyzed. To evaluate the behavior of<br />program, errors entered using the simple scalar simulation software.