A Review of Artificial Intelligence for Games
Xueman Fan
(&) , J. Wu, and L. Tian
Navy Submarine Academy, Qingdao 266199, China
oucfanxm@163.com
Abstract. Artificial Intelligence (AI) has made great progress in recent years,
and it is unlikely to become less important in the future. Besides, it would also
be an understatement that the game has greatly promoted the development of AI.
Game AI has made a remarkable improvement in about fifteen years. In this
paper, we present an academic perspective of AI for games. A number of basic
AI methods usually used in games are summarized and discussed, such as ad
hoc authoring, tree search, evolutionary computation, and machine learning.
Through analysis, it can be concluded that the current game AI is not smart
enough, which strongly calls for supports coming from new methods and
techniques.
Keywords: Artificial intelligence Á Games Á Ad hoc authoring Á Tree search Á
Evolutionary computation Á Machine learning
1 Introduction
Games are fascinating due to the effort and skills needed from people to complete them.
It is the complexity and interestingness nature of games that make them a desirable
problem for AI. The difficulty of the games lies in that their finite state spaces, such as
the possible strategies for an agent, are usually very large. Furthermore, it is often very
hard to assess the goodness of any game state properly. From a computational complexity perspective, many games are NP-hard, meaning that an algorithm for solving a
particular game may run for a long time.
Game and AI have a long history, and the research on game AI mainly focuses on
the construction of game agents. In academic game AI, we distinguish two main fields
and corresponding research activities: board games and video games. Checkers, Chess
and Go are classic board games, which were conquered by AI scientists in 1994 [1],
1997 [2] and 2016 [3], respectively. Compared with the board games, characterized by
the discrete turn-based mechanics and full state of the game is visible to both players,
video games are more challenging. A notable milestone of AI for video games was
reached in 2014 when the game AI developed by Google DeepMind played several
classic Atari 2600 video games on a superhuman skill level only input from the original
pixel [4]. The real-time strategy (RTS) game such as StarCraft has become the next
biggest challenge [5]. RTS is a sub-genre of strategy games where players need to build
an economy and military power in order to defeat their opponents. From a theoretical
point of view, the main differences between RTS games and traditional board games
such as Chess are [6]:
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 298–303, 2020.
https://doi.org/10.1007/978-981-15-0187-6_34
Xueman Fan
(&) , J. Wu, and L. Tian
Navy Submarine Academy, Qingdao 266199, China
oucfanxm@163.com
Abstract. Artificial Intelligence (AI) has made great progress in recent years,
and it is unlikely to become less important in the future. Besides, it would also
be an understatement that the game has greatly promoted the development of AI.
Game AI has made a remarkable improvement in about fifteen years. In this
paper, we present an academic perspective of AI for games. A number of basic
AI methods usually used in games are summarized and discussed, such as ad
hoc authoring, tree search, evolutionary computation, and machine learning.
Through analysis, it can be concluded that the current game AI is not smart
enough, which strongly calls for supports coming from new methods and
techniques.
Keywords: Artificial intelligence Á Games Á Ad hoc authoring Á Tree search Á
Evolutionary computation Á Machine learning
1 Introduction
Games are fascinating due to the effort and skills needed from people to complete them.
It is the complexity and interestingness nature of games that make them a desirable
problem for AI. The difficulty of the games lies in that their finite state spaces, such as
the possible strategies for an agent, are usually very large. Furthermore, it is often very
hard to assess the goodness of any game state properly. From a computational complexity perspective, many games are NP-hard, meaning that an algorithm for solving a
particular game may run for a long time.
Game and AI have a long history, and the research on game AI mainly focuses on
the construction of game agents. In academic game AI, we distinguish two main fields
and corresponding research activities: board games and video games. Checkers, Chess
and Go are classic board games, which were conquered by AI scientists in 1994 [1],
1997 [2] and 2016 [3], respectively. Compared with the board games, characterized by
the discrete turn-based mechanics and full state of the game is visible to both players,
video games are more challenging. A notable milestone of AI for video games was
reached in 2014 when the game AI developed by Google DeepMind played several
classic Atari 2600 video games on a superhuman skill level only input from the original
pixel [4]. The real-time strategy (RTS) game such as StarCraft has become the next
biggest challenge [5]. RTS is a sub-genre of strategy games where players need to build
an economy and military power in order to defeat their opponents. From a theoretical
point of view, the main differences between RTS games and traditional board games
such as Chess are [6]:
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 298–303, 2020.
https://doi.org/10.1007/978-981-15-0187-6_34