Activity models play a crucial role in the realization of the smart environments concept. No matter what application scenarios are involved, in order to make an environment to respond to its inhabitants’ behavioural needs, user activity models are required to support reasoning over streaming sensor data. In addition, once an activity is recognised, activity models are still required to enable application-level functions, e.g., to predict next action, or to identify irregularities for anomaly detection.
Activity modelling, in particular, activities of daily living (ADL) in a Smart Home, is a challenging task due to their unique characteristics. Yet it is critical for activity modelling to generate complete, accurate and adaptable ADL models. Currently there are two mainstream approaches to modelling ADL activities in a SH. One approach is to learn an individual’s activity models from the individual’s existing behavioural datasets. It is usually based on data mining and machine learning techniques, i.e., using data-intensive probabilistic and statistical methods to create and train activity models. The other approach to activity modelling is to manually define and specify activity models. It is motivated by the observations that most ADLs are daily routines taking place in a specific circumstance of time, location and space with relatively fixed types of objects. By making use of rich prior knowledge and domain heuristics activity models can be created through formal knowledge acquisition and modelling technologies. The second approach is essentially a practice of knowledge engineering and knowledge-based systems. As a result, different knowledge representation formalisms have been used for activity modelling and reasoning with each of them displaying strengths and weaknesses.
This project aims to combine two well-known and accepted knowledge formalisms, i.e., the Description Logic and the Event Calculus, to conceive and develop a more expressive and powerful formalism for activity modelling, representation and inference. Description logic, which receives growing attention and usage within the Semantic Web initiative in the form of ontological modelling, representation and inference, has the advantages of creating generalised and specialised activity models in a hierarchical structure with clear semantic, which can be shared, reused and machine processable. The Event calculus, a well-defined logical formalism for modelling and representing actions/events, has the strengths in modelling temporal information, procedural knowledge and event sequences, which are indispensable in modelling complex activities such as interleaved, parallel and concurrent activities. The project will investigate the combination of the two formalisms, and test and evaluate various features and performance of the proposed formalism. In addition, the developed formalism will be used to model and represent activities for specified research scenarios, for example, complex activity modelling, activity recognition or activity learning.
Chen L, Nugent C D, Mulvenna MD, Finlay DD, Hong X, Poland M., (Dec 2008) "A Logical Framework for Behaviour Reasoning and Assistance in a Smart Home", International Journal of Assistive Robotics and Mechatronics, Vol. 9, No. 4, Pages 20-34
Chen L., Bechkoum K and Clapworhty G. (3/2001), Reconciling Autonomy with Narratives in the Event Calculus. AAAI2001 Spring Symposium Series. Stanford University, USA, AAAI Technical Report SS-01-02, pp20-24
First Supervisor: Chen, L Dr
Second Supervisor: Wang, H Dr
Third Supervisor: Liu, J Dr
Collaboration: This project does not involve collaboration with another establishment