Computationally Adjustable Autonomy

Abstract

Reasoning about autonomy is an integral component of collaboration among computational units of distributed systems. This paper introduces an agent-level algorithm that allows an agent to continuously update its autonomy with respect to recurring asynchronous problems with the aim of system-wide collaboration efficiency. The algorithm is demonstrated in a relevant scenario involving NASA space station-based Personal Satellite Assistants, which can handle dynamic situation management that frustrates global collaboration protocols.