Measuring learning progress in intelligent autonomous robots

摘要

AbstractAI researchers claim to understand some aspect of human intelligence when their model is able to “emulate” it. In the contexts of mobile robots, the ability to go from teleoperation to autonomy should accordingly be treated not simply as a trick for robotics hackers but as an important epistemological and methodological goal. In this article the experience derived from designing locomotion control systems for both teleoperated and autonomous multi‐legged articulated robots is desc