Abstract The quest for autonomy has been a pervasive theme in human culture through-out history. In this paper a general definition of autonomous systems is presented and discussed that leads naturally to the establishment of metrics to measure the level of autonomy of a system. This definition is based on the systems ability to achieve goals under uncertainties and it does not involve the means by which the goals are achieved, such as sensing and feedback. This paper takes the point of view that any autonomous system is a control system, and that to achieve higher levels of autonomy one may need to add methods traditionally developed in areas such as operations research and AI. The work presented here is based on earlier work by the author on functional architectures for autonomous spacecrafts.