MULTI-DOMAIN SEMANTIC MODELS AND REASONING FOR SAFETY-CRITICAL URBAN OPERATIONS

Abstract

This dissertation explores the development of an integrated framework for multi-domain semantic modeling and reasoning, coupled to OptaPlanner, a real-time planning, and optimization tool. The investigation is motivated by abstractions from three safety-critical urban application domains: (a) small urban network operations, (b) airplane taxiing operations, and (c) planning and scheduling for disaster evacuation. From a systems modeling perspective, all three domains share the common interests of decision making supported by high-level situational awareness, and effective planning to avoid schedule conflicts and to assure system’s safety. Our integrated approach uses knowledge-based representation and reasoning: (1) to understand the relationships among the physical entities in each application domain that is complicated by other relevant participating domains, (2) reason semantic graphs with external events, and (3) transform and update semantic graphs in response to these external events for making further decisions. We investigate the usages of temporal, spatial, and graph theories, and understand what role ontologies play in deriving appropriate semantic models for urban applications.

Semantic modeling and reasoning capabilities in the dissertation work are handled by Apache Jena and Jena Rules; temporal knowledge and reasoning are driven by time ontology and Allen’s Algebra for temporal relations; Spatial knowledge and reasoning are supported by spatial ontology and class that interfaces (AbstractGeometry) with the Java Topology Suite (JTS); The JGraphT is used to handle graph structures and conduct associated graph analyses, which is a Java library describing graph theory data structures and algorithms. Systems integration of these elements is adopted in the Whistle environment, a small scripting language that is able to process complex data types (i.e., physical units and quantities). While the scope of this dissertation is limited to the three case study applications, we expect that the new knowledge will set us on a pathway to assembly and planning for behavior scenarios across a wide range of urban applications.