To meet the system scheduling requirements of open hybrid real-time systems, an open adaptive real-time scheduling framework, called OARtS (open adaptive real-time scheduling), is presented in this paper, which comprises three key components: accept control, scheduling server and adaptive control. To guarantee the schedulability in the open environment, OARtS only accepts the task whose computing bandwidth requirement is no higher than the system's spare one. To schedule multi-constraint tasks, a two-layer scheduling mechanism is introduced. In the mechanism, the scheduling server components provide concurrent scheduling mechanism for multi-constraint tasks, and each of them is assigned to a bandwidth-independent computing bandwidth and has its specific scheduling policy to schedule its own task queue. To adapt to the change in the open environment, the adaptive control tries to tune the real-time service level so as to make full use of the system computing capability; to adapt to the uncertainty of execution time of the soft real-time task, a fuzzy control engine is used to regulate the task's computing bandwidth according to fuzzy rules of the scheduling error so as to eliminate the scheduling error and to get a satisfactory soft real-time performance.