Which Heroes to Pick? Learning to Draft in MOBA Games With Neural Networks and Tree Search

Sheng, Chen; Zhu, Menghui; Ye, Deheng; Zhang, Weinan; Fu, Qiang; Yang, Wei

doi:10.1109/tg.2021.3095264

Cited by 12 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…XLand [20] also focuses on the generalization capability of agents and supports multi-agent scenarios, but it is not open-source. Existing Interest This environment has been used as a testbed for RL in research competitions 2 and many researchers have conducted experiments under the environment of Honor of Kings [3,4,11,24,25,26,28,29,27].Though some of them verified the feasibility of reinforcement learning in tackling the game [11,26,28,29], they are more focused on methodological novelty in planning, treesearching, etc. Unlike these papers, this paper focuses on making the environment open-accessible and providing benchmarking results, which could serve as a reference and foundation for future research.…”

Section: Motivations and Related Workmentioning

confidence: 99%

Honor of Kings Arena: an Environment for Generalization in Competitive Reinforcement Learning

Wang¹,

Chen²,

Ji³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

This paper introduces Honor of Kings Arena, a reinforcement learning (RL) environment based on Honor of Kings, one of the world's most popular games at present. Compared to other environments studied in most previous work, ours presents new generalization challenges for competitive reinforcement learning. It is a multiagent problem with one agent competing against its opponent; and it requires the generalization ability as it has diverse targets to control and diverse opponents to compete with. We describe the observation, action, and reward specifications for the Honor of Kings domain and provide an open-source Python-based interface for communicating with the game engine. We provide twenty target heroes with a variety of tasks in Honor of Kings Arena and present initial baseline results for RL-based methods with feasible computing resources. Finally, we showcase the generalization challenges imposed by Honor of Kings Arena and possible remedies to the challenges. All of the software, including the environment-class, are publicly available at: https://github.com/tencent-ailab/hok_env. The documentation is available at: https://aiarena.tencent.com/hok/doc/.

show abstract

Section: Motivations and Related Workmentioning

confidence: 99%

Honor of Kings Arena: an Environment for Generalization in Competitive Reinforcement Learning

Wang¹,

Chen²,

Ji³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…So, in this paper, those works are summarized and are put into four categories: Item recommendation, Draft recommendation, Outcome prediction and Strategy making, from Section 2.1 to Section 2.4 respectively, and Section 6 is the conclusion. As Sheng Chen et al [2]'s research, they built their method according to leveraging neural networks and Monte-Carlo tree search (henceforth referred to as MCTS), they used MCTS with a value network and policy network, the value network analyzes the value of current game state while the policy network samples actions for next draft. They tested their model on two datasets individually, based on both matches played by AI and human.…”

Section: Introductionmentioning

confidence: 99%

“…Draft recommendation is literally helping the players to select the most suitable champions in the current game. As shown in Figure 1, champions in the upper left and upper right corners in the figure are the champions banned by both allied team and enemy team, while left and right sides are the champions picked.As Sheng Chen et al[2]'s research, they built their method according to leveraging neural networks and Monte-Carlo tree search (henceforth referred to as MCTS), they used MCTS with a value network and policy network, the value network analyzes the value of current game state while the policy network samples actions for next draft. They tested their model on two datasets individually, based on both matches played by AI and human.…”

mentioning

confidence: 99%

Deep learning in multiplayer online battle arena games

2024

ACE

View full text Add to dashboard Cite

Multiplayer Online Battle Arena games, known as MOBA in abbreviation, are developing rapidly, and more and more new players are growing interests to it. But some parts of these games are quite complicate for those beginners, such as how to pick appropriate champions, how to choose suitable items for purchasing, what is the win rate for current game session and how to make correct strategy decisions. This paper summarized some works, that can help players to solve those complicate parts and understand the game well, using machine learning and deep learning models. These works have all proved their feasibility according to either their result comparing with other baseline methods, or simulating some game sessions played by or against AI using their champion picking, item purchasing and strategy making suggestions. There are also some limitations of these works and some improvements of using machine learning and deep learning in MOBA game industry mentioned in this paper.

show abstract