Robots with high levels of autonomy are crucial for a fast and fruitful development of space exploration, but hard challenges exist.
One of the most challenging goals is to achieve systems capable of predicting accurately the results of their actions, deliver and autonomously develope theories which can be then applied as general rules similar cases.
Towards this direction, the ongoing research focusses on the design, development and integration of a simulation running on the robotic system, capable of dynamically represent the environment and the robot itself including its software modules.
This representation improves the capacity of the system to predict the results of its actions and opens the doors to better decision taking based on knowledge acquired through internal experimentation.
The proposed architecture design and the assotiated methodology cover three life stages of a robotic system: (1) Initial configuration space illumination, (2) Adaptation of internal models to match experience, and of the behaviors that are based on them, and (3) the Accurate identification of capabilities based on real and simulated experiences.
Although still ongoing development, results available from experiments performed on simulations and on real systems in space analog scenarios like a moon crater and lava tubes will be presented.