This paper presents a hybrid AI framework for rescheduling tasks within UAM vertiports. This scheduling challenge is approached as a resource-constrained project scheduling problem (RCPSP), typically solved via mixed-integer linear programming (MILP). However, unlike ideal models, real-world UAM operations are messy, and operator requests are frequently ambiguous. To handle this uncertainty, the proposed framework pairs a Bayesian network to infer intent via dialogue with Answer Set Programming (ASP) to categorize specific ambiguity types. Once the input is clarified, the system generates new MILP constraints and recalculates the schedule, allowing the operator to instantly verify changes against the initial plan.
Kim et al. (Tue,) studied this question.