Learning-Based Attitude Tracking Control With High-Performance Parameter Estimation

This article aims to handle the optimal attitude tracking control tasks for rigid bodies via a reinforcement-learning-based scheme, in which constrained parameter estimator is designed compensate system uncertainties accurately. guarantees exponential convergence of estimation errors and can strictly keep all instant estimates always within predetermined bounds. Based on it, critic-only adaptive dynamic programming (ADP) strategy proposed learn policy with respect user-defined cost function. The matching condition reference signals, commonly employed relevant ADP design, not required scheme. We prove uniform ultimate boundedness critic weight’s under finite excitation conditions by Lyapunov-based analysis. Moreover, an easy-to-implement initial trigger real-time learning process. effectiveness advantages method are verified both numerical simulations hardware-in-the-loop experimental tests.