Stable Grounded Inference in Flexible Resource Scheduling

Systematic temporally flexible solving has proved useful in scheduling with temporal and resource constraints. However, determining resource flaws in a flexible framework is a complex task even for activities with instantaneous resource impacts, and becomes more daunting for linear or more complex impacts. Plan stability is also important for many applications; as new constraints arise, the changes to an existing plan to accommodate them should be minimized to a degree consistent with rapid solving. In this paper we present a method that uses an evolving grounded solution to guide a flexible resource solving process. To promote stability, the successive grounded solutions satisfy a minimal perturbation criterion. We also discuss an encoding of state constraints as numeric resource transactions and some theoretical implications for state reasoning.