Effects of evolutionary configuration of reinforcement learning applied to airship control

The purpose of this Study is to acquire an adaptive control policy for a small airship in a real environment based on reinforcement learning combined with evolutionary construction of the structure. Although reinforcement learning methods could be expected to control a small airship, the unstable property of the airship prevents the learning methods from achieving control of it. To prepare effective state space segmentation for control of an airship, we propose using evolutionary segmentation methods for the reinforcement learning methods. The state space is constructed evolutionary in the simulation environment. Then the acquired state space is applied to learning in the real environment.