Self-Learning Optimal Control with Performance Analysis using Event-Triggered Adaptive Dynamic Programming

Abstract

In this paper, a self-learning event-triggered algorithm is proposed to solve optimal control problems for discrete-time systems. Not quite the same as the time-triggered algorithm, the developed controller is only updated when events occur. To obtain a desired performance, we develop novel triggering conditions with adjustable parameters. Therefore, a near-optimal performance can be obtained as required. Meanwhile, the system is proved to be asymptotically stable based on the developed algorithm. Besides, two neural networks are utilized to implement the algorithm. Furthermore, we test the developed algorithm with different parameter settings, and the impact of parameter settings on the performance is illustrated. Finally, we test both the time-triggered and event-triggered algorithms for comparison, which demonstrates that we obtain comparable performance with reduced computing cost.