Enhancing Pokémon VGC Player Performance: Intelligent Agents Through Deep Reinforcement Learning and Neuroevolution
摘要
Competitive Pokémon battles demand deep strategic thinking and real-time decision-making, posing challenges for players seeking to optimize their performance. This paper presents a novel approach utilizing Deep Reinforcement Learning and Neuroevolution to develop AI agents that excel in the complex Video Game Championships (VGC) format. We address the limitations of existing research by focusing on VGC double battles and their evolving rules. Our intelligent agents learn through game experience, offering competitive challenges to human players and suggesting optimal moves in an intuitive interface. Additionally, we present a tool that leverages the agent’s knowledge to promote the use of underutilized Pokémon, fostering gameplay diversity and enriching the experience for both novice and experienced players. This research demonstrates the potential of AI-powered assistance in enhancing player experience and strategic decision-making within complex competitive games.