In this paper, we present a prototype calendar management assistant that uses fuzzy logic for the representation and satisfaction of user preferences. The system uses ideas from fuzzy scheduling and more loosely from "granular computing" to schedule actions at two different levels of temporal granularity: the task level and the activity level. Tasks are specific actions that have a set start time and are performed throughout a single timeslot (such as a meeting), whereas activities are processes that may not have a fixed start time and which may be spread over many time periods (such as writing a research paper). Users specify start times, deadlines and durations using terms such as morning, late afternoon and around 1 hour which are interpreted as fuzzy functions: the scheduler must produce a valid schedule that maximizes the overall satisfaction of the user's preferences. The present version of the system is also able to schedule plans of tasks or activities (networks of tasks or of activities ordered into an acyclic graph) in which fuzzy constraints between actions may be specified.
展开▼