Designing a controller that works: using formal techniques in robotic systems

The size and complexity of robot controllers is such that it is impossible to predict their performance by conventional means. Debugging and system maintenance are major problems which have only partial solutions. In conjunction with careful structuring, the authors have used formal mathematical techniques (often known generically as formal methods) in designing the software architecture of a mobile robot controller to gain greater understanding of the system and to validate its expected performance. The impetus for this work is a major mobile robot project to provide sensory control in a factory application. In this article, the authors assess the place of formal methods in the cycle of system development