Robotic self-replication, self-diagnosis, and self-repair: Probabilistic considerations

Abstract

This paper reviews issues related to two topics: (1) "self-replicating robots," i.e., robots that can make copies of themselves from spare parts; and (2) robotic teams that can "self-diagnose," which is an essential attribute that will lead to the ability of robots to autonomously and cooperatively repair other robots. In order to quantify the robustness of such robots, statistical measures of the degree of environmental uncertainty that they can handle must be computed. The entropy of the set of all possible arrangements (or configurations) of spare parts in the environment is just such a measure. In the team diagnosis and repair scenario, a team of robots works cooperatively to perform tasks. During this process, all members of the team diagnose each other to determine if there is any abnormal behavior in any team member. When there is, the members of the team can, in principle, cooperate to either repair the malfunctioning team member or disassemble it for use as spare parts. The algorithmic side of this problem is also statistical in nature, and is briefly reviewed here.