Testing control software using a genetic algorithm

Complex software is difficult to test. When that software has been developed by a third party in response to a requirements specification and is to be used in an electronic control unit in the automotive, aerospace or marine industries, this testing process can be even more difficult, but is an essential task. However, testing all possible combinations of inputs to software can be time-consuming, tedious and may be intractable. This paper presents a genetic algorithm (GA) designed to search for significant input and output combinations to a software control system. By ''significant'' is meant those which produce an output (or result) which is not in line with the original specification. It is intended that such a tool should be used to support the human tester by focusing their attention on areas of concern which they can investigate further.