ASR: ASP-based stream reasoning for semantic sensor data processing and activity recognition

Submission note: A thesis submitted in total fulfilment of the requirements for the degree of Doctor of Philosophy to the School of Engineering and Mathematical Sciences, Faculty of Science, Technology and Engineering, La Trobe University, Bundoora.

This thesis proposes the ASR language (Answer set programming (ASP)-based Stream Reasoning language), as a dialect of the dlv language, for stream reasoning with abilities to deal with ambiguous situations and to integrate with Artificial Neural Network (ANN) from machine learning. The ASR language is a dialect of a well-developed logic-based reasoning engine. We equip ASR with stream reasoning functionalities such as handling data stream, performing reasoning continuously, and reasoning about ambiguous situations. With the focus on sensor applications and activity recognition, we chose the ASP technique with dlv language (a state-of-the art ASP solver) to be the foundation for developing ASR. We also utilized ANN, due to its versatility, as an external reasoning technique to be integrated into ASR systems to enhance their reasoning ability. To reason about ambiguous situations, especially in dealing with ambiguous sensor data, we propose a method which uses weak constraints of ASP in combination with the window slide technique and activity relationships. We develop the ASR framework to examine the feasibility, performance, accuracy and practicality of our approach. The ASR framework also allows ASR systems to integrate with ANN seamlessly. As a proof-of-concept, we implemented a case study (“AutoDiary”) which automatically detects a user’s daily activities, and then, provides statistical information about the users’ daily routine patterns and consequently gives alerts and recommendations to the user.