심층 강화학습 기반 골프 전략 에이전트 개발
Golf Strategy Agent using Deep Reinforcement Learning
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This paper presents a golf strategy agent that uses deep reinforcement learning to discover the optimal golf strategy. Here, we have defined the state and the action that are suitable for golf strategy. The state variables comprise an embedded image that incorporates information about the golf course environment, the ball's position, and the hole's direction, as well as the distance to the hole. The action is defined as being in a hybrid action space that combines the chosen golf club ID with the direction of impact. To overcome the challenges of collecting data for deep reinforcement learning, we developed a golf skill model and leveraged it to generate training episodes. We verified the performance of the model developed through the experiment.
Abstract
This paper presents a golf strategy agent that uses deep reinforcement learning to discover the optimal golf strategy. Here, we have defined the state and the action that are suitable for golf strategy. The state variables comprise an embedded image that incorporates information about the golf course environment, the ball's position, and the hole's direction, as well as the distance to the hole. The action is defined as being in a hybrid action space that combines the chosen golf club ID with the direction of impact. To overcome the challenges of collecting data for deep reinforcement learning, we developed a golf skill model and leveraged it to generate training episodes. We verified the performance of the model developed through the experiment.