The NIST Real-time Control System (RCS): an approach to intelligent systems research

摘要

Differences between the deliberative and reactive aspects of behaviour have led to different scientific approaches that have divided the field of cognitive research for at least a century. In the tradition of the behaviourist school of psychology, many present day researchers in advanced robotics and artificial intelligence have abandoned inquiry into the complexities of cognitive behaviour in favour of a minimalist philosophy that focuses primarily on stimulus-response reactivity. In contrast, the Real-time Control System (RCS) developed at NIST and elsewhere over the past two decades provides a model for bridging the gap between deliberative and reactive behaviours. RCS is a reference model architecture for intelligent systems design that consists of a hierarchically layered set of processing nodes. In each node, there are both cognitive and reactive elements. At each layer, entities are recognized, tasks are deliberatively planned, and feedback from sensors closes a reactive control loop. RCS thus integrates and distributes deliberative and reactive functions throughout the entire hierarchical architecture, at all levels and time frames. A comparison is made between RCS and subsumption and some illustrative examples of RCS applications are given.