Tag
This paper presents a Mahalanobis-guided latent out-of-distribution detection method using a VAE to switch between a reinforcement learning controller and an extremum seeking controller in time-varying systems, validated in particle accelerator control.