We present an approach to computing cyclic schedules online and in real time, while attempting to maximize a quality-of-service metric. The motivation is the detection of RF emitters using a schedule that controls the scanning of disjoint frequency bands. The problem is NP-hard, but it exhibits a so-called phase transition that can be exploited to rapidly find a “good enough” schedule. Our approach relies on a graph-based schedule-construction algorithm. Selecting the input to this algorithm in the phase-transition region ensures, with high probability, that a schedule will be found quickly, and gives a lower bound on the quality of service this schedule will achieve.