Abstract: Dimensionality reduction using Variational Autoencoder (VAE) is widely employed in learning diverse state representations, such as in autonomous driving tasks. Conventional VAE-based ...