Abstract : Our method in autonomy evaluation is to mesure robot performances in respect with task complexity. The task is fixed with a navigation task. The complexity changes due to environment and due to the quantity of information the robot have about this environment. Our asumption follows: the more complexe the environment is the less the robot performances are, furthermore the more the robot got information the more the chances it have to make the right decision (wich should conduct to better performances) are. We have designed metrics to measure both the world complexity and information quantity. The mission analyse is made in a global way and in a local way.