Multiple-Goal Recognition from Low-Level Signals

Researchers and practitioners from both the artificial intelligence and pervasive computing communities have been paying increasing attention to the task of inferring users’ high-level goals from low-level sensor readings. A common assumption made by most approaches is that a user either has a single goal in mind, or achieves several goals sequentially. However, in real-world environments, a user often has multiple goals that are concurrently carried out, and a single action can serve as a common step towards multiple goals. In this paper, we formulate the multiple-goal recognition problem and exemplify it in an indoor environment where an RFbased wireless network is available. We propose a goalrecognition algorithm based on a dynamic model set and show how goal models evolve over time based on pre-defined states. Experiments with real data demonstrate that our method can accurately and efficiently recognize multiple interleaving goals in a user’s trace.