Neuro-Optimal Event-Triggered Impulsive Control for Stochastic Systems via ADP

Area of Science:

• Control Theory
• Artificial Intelligence
• Stochastic Systems

Background:

• Traditional control methods often require periodic updates, leading to high computational and communication burdens.
• Event-triggered control strategies aim to reduce this burden by updating only when specific events occur.
• Impulsive control introduces discrete control actions at specific time points, which can be challenging to optimize in stochastic systems.

Purpose of the Study:

• To develop a novel neural-network-based optimal event-triggered impulsive control method.
• To address optimization problems in stochastic systems with event-triggered impulsive controls.
• To reduce the computational and communication burden associated with controller updates.

Main Methods:

• Construction of a general-event-based impulsive transition matrix (GITM) to characterize state evolution across impulsive actions.
• Development of the event-triggered impulsive adaptive dynamic programming (ETIADP) and its high-efficiency version (HEIADP) algorithms.
• Analysis of admissibility, monotonicity, and optimality properties to establish neural network approximation error bounds.

Main Results:

• The proposed ETIADP and HEIADP algorithms effectively solve optimization problems for stochastic systems with event-triggered impulsive controls.
• The event-triggered approach significantly reduces computational and communication overhead compared to periodic control.
• The HEIADP algorithm optimizes multiprocessor system (MPS) resource utilization and reduces memory requirements.

Conclusions:

• The developed neural-network-based control methods provide an efficient and optimal solution for stochastic systems with event-triggered impulsive controls.
• The algorithms ensure that iterative value functions converge to an optimal neighborhood.
• The HEIADP algorithm offers enhanced performance and reduced resource demands, particularly for multiprocessor systems.