Many applications use contexts to understand their environments and make adaptation. However, contexts are often inaccurate or even conflicting with each other (a.k.a. context inconsistency). To prevent applications from behaving abnormally or even failing, one promising approach is to deploy constraint checking to detect context inconsistencies. A variety of constraint checking techniques have been proposed, based on different incremental or parallel mechanisms for the efficiency. They are commonly deployed with the strategy that schedules constraint checking immediately upon context changes. This assures no missed inconsistency, but also limits the detection efficiency. One may break the limit by grouping context changes for checking together, but this can cause severe inconsistency missing problem (up to 79.2 percent). In this article, we propose a novel strategy GEAS to isolate latent interferences among context changes and schedule constraint checking with adaptive group sizes. This makes GEAS not only improve the detection efficiency, but also assure no missed inconsistency with theoretical guarantee. We experimentally evaluated GEAS with large-volume real-world context data. The results show that GEAS achieved significant efficiency gains for context inconsistency detection by 38.8-566.7 percent (or 1.4x-6.7x). When enhanced with an extended change-cancellation optimization, the gains were up to 2,755.9 percent (or 28.6x).

Context-aware applications adapt their behavior based on collected contexts. However, contexts can be inaccurate due to sensing noise, which might cause applications to misbehave. One promising approach is to check contexts against consistency constraints at runtime, so as to detect context inconsistencies for applications and resolve them in time. The checking is typically immediate upon each collected context change. Such a scheduling strategy is intuitive for avoiding missing context inconsistencies in the detection, but may cause low-efficiency problems for heavy-workload checking scenarios, even if equipped with existing incremental or parallel constraint checking techniques. One may choose to check contexts in a batch way to increase the efficiency by reducing the number of constraint checking. However, this can easily cause missed context inconsistencies, denying the purpose of inconsistency detection. In this paper, we propose a novel scheduling strategy GEAS of two nice properties: (1) adaptively tuning the batch window to avoid missing any context inconsistency; (2) generic to checking techniques with no or little adjustment. We experimentally evaluated GEAS against the immediate strategy with existing constraint checking techniques. The experimental results show that GEAS achieved 143-645% efficiency improvement without missing any context inconsistency, while alternatives caused 39.2-65.3% loss of detected context inconsistencies.