Cognitive Products: System Architecture and Operational Principles

摘要

Future commercial products and product assemblies could greatly benefit from recent developments in machine learning, providing the foundation of cognitive products equipped with sensors and actuators and embedded into tangible objects operating in the real world. This paper identifies key challenges in the related fields and provides motivation for further advancements particularly in the domain of resource-constrained distributed and embedded Artificial Intelligence. Enabling cognitive capabilities, such as perception, reasoning, learning and planning, could result in higher reliability, adaptivity and improved performance, however it would require an increased involvement of nontechnical disciplines like cognitive neuroscience. We propose a generic top-level cognitive architecture providing a reference to various research areas involved in this multifaceted field. Conceptual prototypes of two cognitive products, targeting real-world industrial environments, are presented and discussed.