Action selection in an autonomous agent with a hierarchical distributed reactive planning architecture

Abstract

Action selection is a central issue for robotic autonomous agent design. The action selection mechanism used in an autonomous robot needs to be reactive, but must include planning capacities. It should ennble the coexistence of antinomic goals and constraints, and induce a sufficient adaptivity to cope with circumstances unexpected at design time. To our knowledge, no such mechanism embedding all these properties has been implemented in a robotic application.