A Study on Application of World Model in Industrial Drone Systems

Abstract

To construct an autonomous industrial drone system, we implement an agent with the world model in our 3D simulator, and study its performance of the navigation task in a maze environment. We show that the agent can improve its performance through learning. It suggests potential applicability of the world model to real world problems. We discuss possible extensions of the world model and difficulties in applying the world model to real world problems.