Abilities and Capabilities are different concepts. Classical challenges in robotic arise between theoretical computations and realistic efficient feedbacks. We propose a software platform, trying to split the necessary level of (1) actions (abilities), (2) perceptions of sequences of actions, (3) diagnosis and (4) supervision in order to introduce more reactivity and bring together Artificial Intelligence and Autonomous Robotics. Our first experiments will cope with cognitive mapping representations and necessary adaptations needed when inconsistent landmarks are introduced.