Robot control architecture provides structure and principles for designing a variety of specific robot controllers. In general, while there are an infinite number of possible controllers, to date they have been classified into the following basic types of architectures: reactive, deliberative, hybrid, and behavior-based. Interestingly, these architecture types are inspired by biological systems, ranging from insect locomotion to human cognition. We introduce and describe each in turn, then focus on behavior-based control and its biomimetic properties.
An appropriate slogan for reactive control is “Don’t think, react!” Reactive control is directly inspired by the tight coupling between sensory inputs and motor outputs in biological systems, which enable animals to respond to a variety of changing conditions, threats, and opportunities. Similar capabilities are crucial for robots, even for mundane tasks such as navigating around corridors in a populated building. In general, the ability to respond effectively to quickly changing and unstructured environments is termed reactivity, and it typically implies an underlying reactive controller. Reactive responses are inspired by reflexive reactions in animals: consider the ability to pull the hand away quickly from a hot stove or jerk a limb in response to impact. In order to react so quickly, however, a robot cannot take the time to do complex processing. Consequently, reactive systems are defined by a lack of internal models and representations, and, subsequently, a lack of ability to learn or change over time. They are capable of maintaining a short-term state that can allow them to avoid repetitive actions and cycles (e.g., getting stuck in corners), but such an ephemeral state is purposefully lightweight in terms of processing, so as to not slow down the response time of the system.
Online access to SPIE eBooks is limited to subscribing institutions.