Covariance-Driven Self-Triggered Optimistic Planning for the Networked Control of Nonlinear Systems

Networked control systems offer significant advantages in modern control applications, but also face challenges due to communication constraints and noise. This paper introduces Σ-STOP, a Self-Triggered extension of Optimistic Planning tailored for the optimal control of nonlinear networked control systems with noisy dynamics and measurements. Σ-STOP predicts the growth of state covariance along planned control sequences, adding the covariance magnitude and loop closure penalties to the original cost to reach an overall reward. Via near-optimally maximizing cumulative rewards, the method adaptively decides when to close the control loop and rerun the planner, reducing communication frequency under a managed tradeoff with control quality. Simulation results on linear and nonlinear systems demonstrate that Σ-STOP maintains high control performance while substantially decreasing the number of required loop closures.