Cognitive Patterns: An Architecture for Distributed Control of ChemBots

A key challenge to the future deployment of DARPA’s Chemical Robots (ChemBots) ChemBots is a system for controlling them that operates in a distributed fashion. There are three primary requirements: 1) its distributed nature must be transparent to the user; 2) it must exploit local acting and sensing capabilities while maintaining a global common operational picture; and 3) it must be robust to attrition. We propose to use this insight to create an agent architecture called Cognitive Patterns that integrates top-down and bottom-up approaches in a way that is very natural for humans yet very useful for ChemBots. Cognitive Patterns take a formal approach to describing the top-down information that humans use to understand and act in the world, and to combining that information with information in the sensed world in such a way that all participants, automated or human, share context. The use of single Cognitive Patterns for distributed control, as well as the use of libraries of Cognitive Patterns and the real-time, interactive or automated creation of new Cognitive Patterns, will be explored.