Efficient and accurate path planning for mobile robots in complex environments is a crucial aspect of enabling intelligent systems such as autonomous vehicles and wheeled mobile robots. However, the inherent challenges posed by uncertain and evolving landscapes present a critical problem, necessitating innovative solutions that transcend the limitations of traditional planning methods. This paper introduces an avant-garde approach that elevates optimal path planning through the integration of a Hierarchical Conditional Variational Autoencoder (HCVAE). By harnessing the hierarchical architecture of HCVAE, we transcend traditional limitations by endowing our method with unparalleled capabilities in generating non-uniform sampling distributions. This innovation significantly enhances planning efficiency and success rates, all while staying grounded in the fundamental principles of conventional methods. Furthermore, we introduce pioneering improvements aimed at reducing map complexity and accelerating execution, thus enhancing adaptability to unique operational scenarios. We support our assertions with robust empirical evidence derived from a thorough and rigorous set of experiments, highlighting the significant advancements made in the field of robotics and autonomous systems. Our framework based on HCVAE marks a groundbreaking advancement, ushering in an era of unparalleled efficiency and effectiveness in mobile robot path planning.