Reduced Context Consistency Diagrams for Resolving Inconsistent Data

The ability of pervasive context-aware systems to perform efficiently relies on their ability to gather full and unambiguous information about the environment. But raw data collected from sensors is often noisy, imprecise and corrupted, which leads to inconsistencies and conflicts in gathered data. Also environment is only partially observable, thus allowing ambiguities in the knowledge about its state. In the paper we present reduced context consistency diagrams (RCCD), data structures that allow to store the information about the environment even with the presence of inconsistencies, conflicts, and ambiguities. We provide a mechanism for reasoning about the current situation using these diagrams, and show how to obtain information about the most probable situation at each moment of time. The case study shows the 50% reduction in incorrect sensor readings. The evaluation shows RCCD to be applicable to real-time context inference problems.