Unsupervised Reinforcement Learning (URL) allows agents to develop general behaviors and representations without relying on external rewards. This survey provides a clear overview of the field, organizing recent advances through a unified framework along four axes: learning paradigm, optimization objective, data regime, and agent quantity. We look at types that range from exploration driven by intrinsic rewards to discovering skills based on competence. We classify optimization goals, including maximizing entropy and mutual information, as well as maintaining self-supervised consistency. We also review data regimes that include exploration from scratch to offline datasets. By tracking the field’s development from early curiosity-based methods to recent foundation-model-driven pretraining, we highlight new trends focused on data-driven, scalable designs. Finally, we discuss important challenges, such as long-term exploration, transferring skills from simulation to real-world settings, and open-ended learning, while outlining a path for future research on developing autonomous agents.