ABSTRACT
This chapter considers the possibility of achieving autonomy in a robotic system, the ability to function competently over an extended period without human intervention. It focuses on the scientific and engineering prerequisites and consequences of the aim, leaving aside, for the most part, the important social, economic, and philosophical consequences. The chapter explains with M. Brady and R. Paul's neat characterisation because it encapsulates both the dream of autonomous robots and the essential ingredients for its practical realisation. It explores one of the problems that besets designers of autonomous robotic systems using a research project currently in progress at Edinburgh as a pointer to what might eventually be possible. SOMASS incorporates both conventional symbolic reasoning and tacit knowledge in the form of skills or behaviours providing a nontrivial agent for the planner to instruct. However, the reasonableness of the performance viewpoint disguises the essential difficulty with autonomous robotics: The robots are artifacts, and artifacts must be designed.
