Curling Strategy Teaching Case Design Based on Deep Reinforcement Learning

Reinforcement learning is becoming more and more important in the field of artificial intelligence. Compared with traditional supervised learning and unsupervised learning, reinforcement learning has wider application scenarios and higher flexibility. As a reinforcement learning problem in continuous space, curling has a good representative significance in the field of reinforcement learning, students can have a deeper understanding of the principles and applications of deep learning and reinforcement learning, thinking deeply about the differences and relationships between artificial intelligence and human intelligence. The reinforcement learning algorithm in curling uses the Monte Carlo tree search method, which covers many basic theoretical knowledge of computer science, such as machine learning, mathematics, algorithms, and data structures. In this paper, we propose a case study of reinforcement learning teaching based on digital curling. First, we designed and implemented a digital curling simulator and made a visual interface. Then, to make the algorithm more understandable, we implement a visual explanation of Monte Carlo search trees. Additionally, we have also provided the digital curling competition mode, so that students can realize the competition of different algorithms. Students can effectively improve their understanding of reinforcement learning by using their own algorithms to fight against digital curling.