Methodologies for the Development of Qualitative Spatial and Temporal Reasoning Applications

Abstract

A great variety of scientific, engineering-based, and commercial application domains are fundamentally grounded in concepts of space and time. Over the last three decades there has been significant interest in more human-focused and intuitive qualitative spatial and temporal reasoning (QSTR) methods, which address the inherent limitations of purely numerical approaches for reasoning about space and time. However, despite the extremely significant theoretical advances that have been made in the QSTR field, there is a distinct absence of commercial and industrial applications that utilise QSTR calculi. The central issue is that relatively little research has addressed the unique challenges of designing and developing QSTR-based applications in comparison to more traditional systems that employ numerical processing techniques. The primary objective of this thesis is to support software engineering practitioners in the development of applications that utilise QSTR calculi. Five QSTR application case studies, which cover a range of diverse application domains, are presented and analysed throughout this thesis to motivate the development of effective methodologies. Furthermore, a comprehensive definition of QSTR applications is presented to provide a formal basis for establishing methodologies that address three major areas of QSTR application development: requirements specification, design, and validation. Design methodologies are presented that enable developers to evaluate the efficacy of numerous QSTR calculi with respect to QSTR application functional requirements. Additionally, the design methodologies adapt object-oriented concepts and machine learning techniques to facilitate the development of custom, high level, application-specific qualitative relations and constraints. Four key validation methodologies are adapted from well known techniques in software engineering: unit testing, integration testing, test coverage, and mutation testing. Furthermore, a novel metric called H-complexity is presented and used to define four additional test coverage classes that a developer can employ to assess the efficacy of a test suite. Finally, a meta-validation methodology is established that enables developers and the QSTR community to empirically investigate the efficacy of QSTR validation techniques. Experiments are conducted using the meta-validation methodology and the results are analysed to identify the most effective utilisation of QSTR validation methodologies according to the software development process being employed.