Selecting behaviors using fuzzy logic

Abstract

Behavior-based systems, i.e. systems that use behaviors as a way of decomposing the control policy needed for accomplishing a task, are very useful in making robots cope with the dynamics of real-world environments. However, these systems still need to be extended to design robots having multiple and possibly conflicting goals, requiring planning, and getting more out of their behaviors. One way of doing that is to allow behaviors to be selected dynamically and to blend their actions in order to get more complex behaviors. This paper addresses these issues by presenting a control architecture that, when using fuzzy logic, allows behaviors to be efficiently selected by different sources. A simulated world for mobile robots is used to illustrate these ideas.