4D/RCS A Reference Model Architecture for Intelligent Unmanned Ground Vehicles

摘要

4D/RCS consists of a multi-layered multi-resolutional hierarchy of computational nodes each containing elements of sensory processing (SP), world modeling (WM), value judgment (VJ), and behavior generation (BG). At the lower levels, these elements generate goal-seeking reactive behavior. At higher levels, they enable goal-defining deliberative behavior. Throughout the hierarchy, interaction between SP, WM, VJ, and BG give rise to perception, cognition, imagination, and reasoning. At low levels, range in space and time is short and resolution is high. At high levels, distance and time are long and resolution is low. This enables high-precision fast-action response over short intervals of time and space at low levels, while long-range plans and abstract concepts are being formulated over broad regions of time and space at high levels. 4D/RCS closes feedback loops at every level. SP processes focus attention (i.e., window regions of space or time), group (i.e., segment regions into entities), compute entity attributes, estimate entity state, and assign entities to classes at every level. WM processes maintain a rich and dynamic database of knowledge about the world in the form of images, maps, entities, events, and relationships at every level. Other WM processes use that knowledge to generate estimates and predictions that support perception, reasoning, and planning at every level. 4D/RCS was developed for the Army Research Laboratory Demo Ⅲ program. To date, only the lower levels of the 4D/RCS architecture have been fully implemented, but the results have been extremely positive. It seems clear that the theoretical basis of 4D/RCS is sound and the architecture is capable of being extended to support much higher levels of performance.