This chapter introduces the research on robot learning, with a focus on Robot Online Learning (ROL) frameworks. It begins by defining ROL and motivating its necessity for robots operating in dynamic environments. Two prominent ROL frameworks are then presented: one based on Positive–Negative (P–N) learning and the other leveraging knowledge transfer. A comparative analysis highlights the strengths and weaknesses of each approach. Specifically, while P–N learning operates autonomously, it is susceptible to self-bias. Conversely, knowledge transfer can mitigate self-bias but requires external guidance and must address potential conflicts between internal and external knowledge sources. The chapter further explores strategies for mitigating catastrophic forgetting, a critical challenge in long-term ROL. Finally, it demonstrates how ROL can be applied to enhance socially-compliant robot navigation in extended, cross-environment deployments.