Issues in Providing Adjustable Autonomy In Three Layered Architectures

Pete Bonasso

Three-tiered robot control architectures, such as 3T, Atlantis and the Remote Agent Architecture, have been applied in some incarnation to dozens of field, service, and space robots and other computer controlled machines for almost a decade. In every case, some attention has had to be paid to interacting with the human creators, programmers and users of these robots. In the past five years, since 1994, the developers of 3T at TRACLabs have experienced a variety of needs for adjusting the level of autonomy in the control of these machines in various applications at the Johnson Space Center. Beginning in late 1995, we began a research program to begin to put our ad hoc experiences into a set of principles embodied in adjustable autonomy (AA) software tools for the architecture. We have seen the need for humans to be involved not only at the obvious deliberative layer, but all the way down to teleoperating what were supposed to be machines running autonomously. Moreover, we have found that one guiding principle seems to put all of our AA endeavors in a unified light: design the computer controlled machine and it’s control architecture for full autonomy; then relax the autonomy restriction at each level, beginning with the highest.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.