A systematic agent framework for situated autonomous systems

Abstract

We present a formal framework of an autonomous agent as a collection of coordinated control loops, with a recurring sense, plan, act cycle. Our framework manages the informa-tion flow within the partitioned structure to ensure consis-tency in order to direct the flow of goals and observations in a timely manner. The resulting control structure improves scalability since many details of each controller can be encap-sulated within a single control loop. This partitioned agent design promises a domain-independent, scalable and robust approach for control of real-world autonomous robots oper-ating in dynamic environments. We validate our framework with experimental results from deployments in two different real-world domains.