A Hybrid Approach in the Development of Behavior Based Robotics

Abstract

In this paper we present the development of a method that combines the evolutionary robotics approach with action selection. A collection task is set in an arena where a Khepera robot has to collect cylinders that simulate food. Furthermore, two basic motivations, labeled as 'fear' and 'hunger', both affect the selection of the behavioral repertoire. In this paper we propose an initial evolutionary stage where behavioral modules are designed as separate selectable modules. Next, we use evolution for optimizing the motivated selection network employed for behavioral switching. Finally, we compare evolved selection with hand-coded selection, which offers some interesting results that support the use of a hybrid approach in the development of behavior-based robotics.