Making artifcial life forms do something useful

摘要

To be useful, an autonomous agent (like a mobile robot) needs observe and act in an environment in order to complete a given task. Traditional control architectures for mobile robots relied on accurate world models of the robot's environment. By this, the communication with other robots, humans, computers and within the robot's modules was straightforward - but not situated.rnRecent behavior-based control architectures were implemented to provide fast reactions to a dynamically changing environment. The representations within the agent strongly depend on the interaction with the environment through its sensors and actors. The internal representations are different from those used in a language for inter-robot communication which uses pre-defined fixed symbols.rnGoal-orientedness is the key to usefulness which is not yet provided by current behavior-based control architectures. These behavior-based architectures are reflex-based only and not adaptive. By introducing a memory and judgement system, it is possible to make your goal the robot's desire and achieve real adaptive behaviors.