Frameworks for Semantic Integration of Intelligent Sensor Processing Systems

The problem of unifying symbolic knowledge and concept representations and systems with the sensor inputs from the external environment is fundamental to the development of autonomous sensor-to-symbol systems. Existing approaches do not provide ways for allowing systems to rapidly incorporate new knowledge. The proposed sensor-to-symbol framework addresses the following key capabilities: • Provides a unified approach to representation that encompasses sensor data, features, similarity representations, and classes/categories/types. • Defines an integrated architecture and system engineering approach that focuses on iterative development, adaptation, and knowledge acquisition. • Support for integration of heterogeneous computation models through model-driven system development approaches. The Phase I effort will focus on defining the key technologies, requirements and design approaches of the sensor processing framework. A key innovation of the proposed approach is that it aims to provide a unified knowledge and concept representation across all abstraction layers, paving the way towards semantic integration of large-scale machine intelligence systems. Semantic integration is a promising approach because it facilitates a system’s ability to model and describe itself and hence provides the means for adaptation and learning competencies to interact with the competencies that perform processing and reasoning.