Using Constraint Optimization for Conflict Resolution and Detail Control in Activity Recognition

In Ambient Assisted Living and other environments the problem is to recognize all of user activities. Due to noisy or incomplete information a naive recognition system may report activities that are logically inconsistent with each other, e.g., the user is sleeping on the couch and at the same time is watching TV. In this work, we develop a rule-based recognition system for hierarchically-organized activities that returns only logically consistent scenarios. This is achieved by explicitly formulating conflicts as Weighted Partial MaxSAT clauses to be satisfied. The system also has the ability to adjust the desired level of detail of the scenarios returned. This is accomplished by assigning preferences to clauses of the SAT problem. The system is implemented and evaluated in a real Ambient Intelligence experimental space. It is shown to be robust to the presence of noise; the level of detail can easily be adjusted by the use of two preference parameters.