In the vast realm of AI, robotic capabilities have soared, yet navigating partially unknown environments remains their Achilles’ heel. Robots lack the cognitive finesse to seamlessly transfer tasks across varying domains. This limitation hinders their adaptability and problem-solving abilities. The EU-funded INVERSE project aims to advance robotic cognition and bridge the gap between expectation and execution in unexplored territories. Specifically, it uses continual learning, refining robotic skills through experience and human feedback. By mimicking human learning processes, INVERSE enables robots to understand, act, and predict consequences in diverse domains. Human supervision plays a pivotal role, streamlining the refinement loop for practical deployment. INVERSE’s effectiveness will be showcased in two real-world scenarios.