Robot Controller Architectures for Autonomous and Reactive Robotic Systems
Abstract
Robotic controllers refer to the central computing core of robotic systems, responsible for mechanical motion, sensing, actuation, end-effector operation, and interfacing with the environment in a non-deterministic manner. The full-text study offers a theoretical review of robotic controller architectures including autonomous robots, semi-autonomous robots, and reactive robots. It combines the content presented in the article about robot controllers with modern research regarding deliberative controllers, reactive controllers, hybrid, hierarchical robot control systems, skill-based robotic control systems, soft-robots control, and adaptive robot control systems. The paper firstly introduces the idea of robot control as an integration process of perception, state estimation, planning, motion execution, feedback, and event processing. It further explains record-and-playback programming, open-loop and feed-forward control, closed-loop control, reactive control, artificial intelligence-based control, and interrupt handling. It turns out that the use of open-loop control makes sense when performing well-calibrated and predictable actions. In contrast, it is necessary to apply closed-loop control in cases when it is important to correct the state errors caused by disturbances and uncertainty. Reactive controllers help make decisions based on sensory information quickly, although arbitration schemes are required for the selection of behaviors when they run concurrently.
