AbstractIn this article, a novel self‐triggered mechanism is developed to reduce the computation burden and communication bandwidth for the optimal regulation problem of nonlinear dynamical systems. Compared with existing results, this article can avoid continuous measurement of online signals while achieving the performance optimization with closed‐loop system stability guarantee. The self‐triggered mechanism is combined with the actor‐critic structure for performance optimization, where the critic is trained to provide a guideline to improve the actor. The actor‐critic learning is implemented as a hybrid system, where the critic weights update as a continuous flow, and the actor weights are adapted intermittently. The simulation study is conducted to verify the effectiveness of the proposed self‐triggered actor‐critic learning.