Abstract: Most of real-life scheduling problems are semi on-line. Recently, how to solve such dynamic problems is a hot topic in the research of artificial intelligence. The semi on-line scheduling is introduced and the constraint models are analyzed and categorized. By defining the relevant concept of monotony about constraints that appear in general dynamic constraint models, a kind of constraint extending is formalized. Based on this dynamic constraint modeling, a sound dynamic constraint solving algorithm is designed to deal with the scheduling problems. Finally, an application example of semi on-line discrete resource scheduling problems is given. Experiments show that the algorithm is valid.