A Hybrid Context-aware Middleware for Relevant Information Delivery in Multi-Role and Multi-User Monitoring Systems: An Application to the Building Management Domain
|Kustantaja||Tampere University of Technology|
|Tila||Julkaistu - 1 kesäkuuta 2018|
|Nimi||Tampere University of Technology. Publication|
In theory, the diversity and availability of new data should lead to more informed users and better decisions. In practice, it overwhelms user capacities to perceive all available information and leads to the situations, where important data is hindered and lost, therefore complicating understanding of the ongoing status. Thus, there is a need in development of new solutions, which would reduce the unnecessary information burden to the users of the system, while keeping them well informed with respect to their personal needs and responsibilities.
This dissertation proposes the middleware for relevant information delivery in multi-role and multi-user BMS, which is capable of analysing ongoing situations in the environment and delivering information personalized to specific user needs. The middleware implementation is based on a novel hybrid approach, which involve semantic modelling of the contextual information and fusion of this information with runtime device data by means of Complex Event Processing (CEP). The context model is actively used at the configuration stages of the middleware, which enables flexible redirection of information flows, simplified (re)configuration of the solution, and consideration of additional information at the runtime phases. The CEP utilizes contextual information and enables temporal reasoning support in combination with runtime analysis capabilities, thus processing ongoing data from devices and delivering personalized information flows. In addition, the work proposes classification and combination principles of ongoing system notifications, which further specialize information flows in accordance to user needs and environment status.
The middleware and corresponding principles (e.g. knowledge modelling, classification and combination of ongoing notifications) have been designed contemplating the building management (BM) domain. A set of experiments on real data from rehabilitation facility has been carried out demonstrating applicability of the approach with respect to delivered information and performance considerations. It is expected that with minor modifications the approach has the potential of being adopted for control and monitoring systems of discrete manufacturing domain.