Learning Hierarchical Action Selection for an Autonomous Robot

In this paper we describe an approach of controlling an autonomous robot by means of a hierarchical organised control structure. The realised action selection mechanism is capable of learning to switch between different modes of actions with respect to the internal state of the robot. We present an approach that realises a learning action selection mechanism in a hierarchy of sensory and actuator layers. The sensory values yield the internal states which serve as a basis for the action selection. In addition, the internal states are used to calculate the reinforcement signal that trains, and improves the action selection.