DWELL-TIME GUARD FOR FUNCTIONALLY STABLE STRATEGY SWITCHING IN SOFTWARE CONTROL SYSTEMS

Event-driven control software adapts by switching strategies when monitors report changes. Even if every strategy is stable on its own, switching too often can drive the controlled variables away from the goal and violate functional stability. We embed the classical dwell-time condition into a strategy-selection design pattern as a guard. Each strategy declares its guaranteed decay rate in a contract, an impact matrix bounds how much one switch can worsen the error measure, and the guard forbids switches closer together than the number of steps computed from these two quantities. On a two-strategy benchmark of individually stable but non-normal dynamics, 129 of 150 unguarded runs diverged, whereas every run converged for dwell times of 6 steps or more; the computed bound (10 steps) exceeded the worst-case threshold (7 steps) by a factor of 1.4. A debounce filter calmed random detector flicker but failed under periodic requests, which the guard handled; the two are complementary.