Modeling Hydra Behavior Using Methods Founded in Behavior-Based Robotics

Chapter 1IntroductionDuring the past decades autonomous robots have undergone rapid development, and itis expected that such robots will play an increasingly important role in our society in thefuture [4, 5, 57]. Lawn mower robots, vacuum cleaning robots, and various entertainmentrobots reached the market several years ago, while robots designed for domestic assistance,service, and surveillance are under development. Examples of various autonomousrobots, and their application areas, can be found in [4, 5, 58]. Autonomous robots areexpected to operate in unstructured environments, i.e. environments that change in anunpredictable way. For such robots traditional control strategies defined a priori are ofteninsufficient, since the robot must be able to perceive and adapt to changes in its environmentin order to function successfully.There are many different approaches to the problem of generating brains 1 for autonomousrobots. In classical artificial intelligence (AI) the flow of information is arrangedin a sequential way: the robot uses information acquired through its sensors tobuild a world model, based on which it reasons about the effect of various actions. Asa last step, the robot then decides upon an action, which is executed. This sense-planactapproach of classical AI requires implementation of high-level reasoning, somethingthat turned out to be difficult, as pointed out e.g. in [9, 67]. Thus, while classical AI hasbeen successfully applied in various fields, e.g. pattern recognition, its goal of generatingintelligent artificial agents has not (yet) been reached [67].An alternative to the classical AI approach to machine intelligence, behavior-basedrobotics (BBR) was introduced in the mid 1980s, by R. Brooks and others. See e.g. [8]for one of the pioneering works in BBR. In contrast with classical AI, behavior-basedrobotic brains are designed using a parallel, decentralized architecture. In BBR the aim isto organize simple, often reactive behaviors 2 in such a way to make the robot functionadequately in an unstructured environment. (Introductions to BBR and AI can be found1 Sometimes also referred to as control systems.2 Direct coupling between sensors and actuators.1