Research on Artificial Intelligence Algorithm and

Its Application in Games

NANDHIKA SRI .P 1 , DEVIKA S.Y 2 , Dr. GANESH .D 3

1 Student, School of CSIT , JAIN Deemed-to-be University,Bangalore jpc222145@jainuniversiy.ac.in
2 Student, School of CSIT , JAIN Deemed-to-be University,Bangalore jpc220012055@jainuniversiy.ac.in
3 Professor, School of CSIT , JAIN Deemed-to-be University,Bangalore. d.ganesh@jainuniversity.ac.in

Abstract—With the in-depth development of intelligent their own games [8]. Because game AI has not made great
technology, game artificial intelligence (AI) has become the progress like graphics technology and physical simulation
technical core of improving the playability of a game and the technology, it provides a space for game innovation and
main selling point of game promotion, deepening the game alienation. The application of graphics technology and
experience realm. Modern computer games achieve the realism physical characteristics simulation technology has been
of games by integrating graphics, physics and artificial insufficient, making a game unique [9]. As the technicalcore
intelligence. It is difficult to define the meaning of realistic game of improving game playability and the selling point of many
experience, but generally speaking, it usually refers to the commercial games promotion, game AI gives players the way
immersion of the game and the intelligence of non-player
to generate behavior and emotional interaction with non
characters in the game. As the technical core of improving game
player characters in the game, and promotes the realm of
playability and the selling point of many commercial games,
game artificial intelligence gives players a way to interact with
game experience to a higher level [10]. How to endow the non
non-player characters in the game, and promotes the realm of player characters with credible intelligence, so that they can
game experience to a higher level. Based on this, this paper more truly reflect the human like behavior, emotion, and even
analyzes the history and present situation of artificial self-learning to adapt to the changing game environment, has
intelligence in game development, and puts forward the become a hot spot of game research and development at home
possible changes and impacts of artificial intelligence technology and abroad [11]. This paper analyzes the history and current
based on machine learning on game development in the future. situation of AI in game development, and puts forward the
possible changes and influences of AI technology based on
Keywords—Artificial intelligence, Game experience, Machine machine learning on game development in the future,
learning including intelligent game design, intelligent iteration and
subsequent development strategy generation and execution
I. INTRODUCTION ability, highly intelligent role, dynamic adaptation and
With the rapid development of all kinds of computer game constantly changing game experience.
engine, especially the development of computer game
industry [1]. With the significant improvement of computer II. KEY TECHNOLOGIES OF GAME AI
hardware level, the game engine used by game designers is Artificial intelligence technology is mainly based on
constantly updated. As the basis of game development, game computer. It studies the activity of transforming into artificial
graphics technology is updated every year. The quality of a intelligence through electronic technology. Its overall design
game depends largely on graphics technology [2]. However, shows intelligence and can achieve the work that human
with the increasing perfection of graphicstechnology, people intelligence can accomplish. Now, there is a clear difference
are no longer satisfied with the beautiful and gorgeous audio- between the type of artificial intelligence in commercial
visual experience, but instead pursue deeper game games and the type of artificial intelligence as a player
connotation [3]. Modern computer games achieve the reality playing games. One of the characteristics that distinguishes
of the game by integrating graphics, physics and artificial human beings from animals is the ability of learning and
intelligence (AI) [4]. Realistic game experience is difficult to innovation. Animal learning mostly stays in the stage of
define its connotation, but generally speaking, it usually conditioned reflex, while human learning has developed to
refers to the immersion of the game and the intelligence of the height of multi-dimension, rational speculation and
non player characters in the game [5]. A successful game that creation and invention. Many problems in game development
is popular in the market must have not only gorgeous visual can be solved perfectly by traditional qualitative AI
effects and pleasanthearing, but also highly realistic artificial techniques such as state diagram search and regular script.
intelligence control system [6]. The application of artificial intelligence technology in games
When game developers apply AI to computer or game is mainly to design immersive scenes for gamers, so that
console games, they will make the majority of players feel players can immerse themselves in this relatively real
that the enemy controlled by computer AI system (NPC)they situation. Under normal circumstances, the computing
face has human intelligence just like the real enemy, so that capacity and storage capacity of the general computer cpu are
the players can leave a realistic experience [7]. Game very limited. Using artificial intelligence technology to
developers need to find innovations that further alienate program and calculate the characters in the game, the
interaction between the whole game and the
players is enhanced, and the sense of reality is also increased. according to the characteristics of different players. Taking
The finite state machine can be used to describe the intelligent Chinese chess as an example, the existing AI technology can
behavior of non-player characters. However, when characters only simply control the search depth of the state tree and
need to deal with more states, the finitestate machine will evaluate the complexity of the function [12]. The explosion
become very difficult to understand and maintain because of of state combination caused by finite state machine can be
the explosive growth of scale. alleviated to a certain extent by using hierarchical state
machine, in which the behavior of non-player characters is
In a grid-based system, the essence of path search is to use decomposed into several small tasks, but the hierarchical state
an algorithm to find the shortest path from the current node to machine still uses a large number of state transitions, which
a certain target node. Modern games are using variations of is also difficult to manage. Selection nodes without priority
these technologies. Of course, artificial intelligence in do not judge the conditions from the first child node every
commercial games is much more complicated than this, but time, but from the child node executed last time. If the
most of them are different versions based on these basic conditions are met, the node will continue to be executed, and
principles. From the player's point of view, the advantages of if the conditions are not met, other child nodes will continue
learning are mainly reflected in the fact that the game can to be judged in turn. The actual driving process of NPC
dynamically adjust the difficulty according to the player's motion system is obtained through visual behavior analysis,
level, so that players can get a great experienceat all times, and a unified visual and behavioral model is formed, as shown
or choose different tactics independently in Figure 1.

Figure 1 Vision and behavior model

The artificial intelligence technology in the game should simulations and systems that are closer to realizing the
not be too smart, at least not more than half of the player's complexity of the real world, large studios may challenge
level, otherwise the player will lose interest in the game. the limits. Making games will be much easier, and producers
Therefore, when artificial intelligence technology is applied can make bigger games. These open-world games will become
in games, on the one hand, it should limit its effectiveness and larger, and the game rules system is changeable. The rules are
prevent players from being bruised, on the other hand, it different every time you play, and even between different
should also increase the development and application of players' terminal computers. Artificial intelligence technology
intelligent technology in games. Advanced artificial in games has great potential in future development. Future
intelligence has better original computing ability, or the games can change with the changes of games. Most games
machine has to face the solution thinking, or the like, but in rarely have great repetitions in requirements, but they can be
the game, the designer hardly pays attention to it at all. What used for reference and unified in development models. This
game designers really want is for players to have a good model can be used in most games with complex situation
experience. Usually, when designing games, designers want decision and evaluation. As for the self-learning function of
to design an experience for players. Designers want to know the model, offline learning can be used in game development,
what players will experience when they reach this point in the so as to extract many parameters that have great influence on
game. Therefore, if an artificial intelligence is to be placed, it decision- making, and train neural network, so that the
is expected that the artificial intelligence is predictable. When network can learn the game experts' understanding of the
designing games, game designers need to balance the virtual game to a certain extent. At present, the role of advanced
game world. On the one hand, they should attract players' artificial intelligence is not to act as in-game artificial
interest in games, and on the other hand, they should make intelligence in games, but to replace real players to play games
games challenging. The design of artificial intelligence in and test the games themselves.
games is complicated and difficult, which plays a very
important role in the quality of games. Therefore, to design a When a game goes on the market, the behavior of non-
balanced and difficult game, the application of artificial players will not be arranged in advance, and the longer the
intelligence technology is very important. game is played, the game itself can adjust itself according to
the game progress, so as to better adapt to the player's
III. APPLICATION OF ARTIFICIAL INTELLIGENCE BASED ON progress of the game. Figure 2 is the frame of game artificial
MACHINE LEARNING IN GAME DEVELOPMENT intelligence detection and tracking system using image error
technology.
A. Intelligent game design
When making an open world environment and creating
 decisions, while the input of other non-training sample data
depends on the generalization ability of neural network.
In the game, the most classic problem of path search is the
game maze problem, which is also one of the difficulties in
game design. Many intelligent problems, such as chess game,
strategic decision-making, robot path planning, etc., can be
transformed into finding the optimal path of maze. Figure 3
shows the structure of agent nodes.

Figure 3 The structure of the agent node

With the increase of path density, the shortest path

Figure 2 Game artificial intelligence detection and tracking system between nodes increases, as shown in Figure 4.
framework using image error technology

Game software also serves as a tool for training advanced

artificial intelligence. Artificial intelligence researchers
mainly regard games as a method to measure the intelligence
level of software, because the virtual world with strict rules
and reward system is a particularly useful environment for
training certain software. Traditional breadth-first search and
depth-first search are both non- heuristic. By simulating the
search process with computer images, we can easily prove that
non-heuristic path search is actually an exhaustive method,
which searches the waypoints around the pathfinder in a fixed
order until the destination waypoint is found. Weights will
bring a certain tendency to randomness, but there will still be
no starvation of other child nodes [13]. If the weights of all Figure 4 Path density and node path relationship
child nodes are set to the same value, the selected node
becomes a completely random selected node. Machine Different losses can be set for different terrains, such as
learning is an essential data mining tool for in-game analysis, swamps, hills, stairs, etc., which can set higher moving cost
so game studios can study players' behaviors and decipher than flat ground. Hierarchical path search is necessary when
new insights to improve games over time. In fact, the finite the map is very large or when you want to cross the map to
state machine maintains a state graph, the vertices of the find the path. In the real world, if you want to find a village,
graph are states, the connection between vertices is a if the path is too far, people may not know how to get there
transition from one state to another, and the logic of transition directly, but you can go to the county where the target is
is described by rules. The process of state jump can be located first, then find the way to the township where the
described by behavior tree. The key to the application of target is located, and finally find the destination. The most
advanced artificial intelligence in this field will be a real game exciting element in the prospect of the future may be not only
character driven by artificial intelligence, or an overall game the creative role played by software in the process of
design artificial intelligence system, which can change, grow constructing the art of games, but also the ever-changing
and react like human beings when playing. and tailor-made experience created by this technology, which
will never be boring to deal with only one goal at a time. As
B. Artificial intelligence routing a result, only one action is executed at a time. Moreover, only
It is difficult to find the existing linear formula to describe the currently processed target rules are tested, thus providing
the solution model of the non-qualitative AI problems in a context. Rules are processed in sequence, and the first rule
games, but these problems can be transformed into specific that meets the conditions passes control to another new target.
patterns or nonlinear mappings. A game must have a perfect Automated game design can make players experience many
process to effectively support its smooth running. A simple times, because the game can be redesigned and refreshed
game can jump in a fixed mode after determining the constantly.
established mode, but when encountering non-discrete
conditional decision, conditional statements have limited IV. CONCLUSIONS
ability. Here, we can abstract the experience of human With the rapid development and progress of China's
reading competition into several parameters to train neural economy, great breakthroughs have been made in science and
networks. Let the network learn these representative technology, especially in the field of artificial
 Practice, vol. 53, no. 2, pp. 72-77, 2014.
intelligence, which has great potential for mining. Like
[2] Liu Yifan. Analysis of the application of artificial intelligence in game
today's game graphics engine, artificial intelligence engine development. Digital design. CG WORLD, vol. 8, no. 7, pp. 86- 86,
will become an indispensable part for large-scale game 2019.
development. At present, the traditional in-game artificial [3] Wang Feiyue. Artificial intelligence wins in multi-role games.
intelligence can still fulfill the mission entrusted by the Chinese Science Foundation, vol. 34, no. 2, pp. 85-86, 2020.
developers, and create realistic intelligence illusion by [4] Cao Kunze. Artificial intelligence and its application in the game field.
combining with other audio-visual interactive means. Science and Technology Communication, vol. 257, no. 8, pp. 162-163,
Machine learning is an essential data mining tool for in- game 2020.
analysis, so game studios can study players' behaviors and [5] Zheng Xin, Zhang Jing. Educational games in the era of artificial
intelligence: development opportunities and trends. Digital Education,
decipher new insights to improve games over time. It is vol. 31, no. 1, pp. 33-37, 2020.
difficult to find the existing linear formula to describe the
[6] Wang Yuxuan. Analysis on the development and application of game
solution model of the non-qualitative AI problems in games, artificial intelligence. Science and Technology Communication, vol.
but these problems can be transformed into specific patterns 11, no. 2, pp. 141-142, 2019.
or nonlinear mappings. As for the self-learning function of the [7] Li Kun, Li Ping, Li Libo. Design and Implementation of MOBA
model, offline learning can be used in game development, so Game Artificial Intelligence. Computer and Information Technology,
as to extract many parameters that have great influence on vol. 154, no. 4, pp. 12-15, 2018.
decision-making, and train neural network, so that the [8] Wang Shiying. The simplified version of "artificial intelligence" is
network can learn the game experts' understanding of the rejected in management. School of Business, vol. 1, no. 1, pp. 28-28,
2015.
game to a certain extent. After thelatest artificial intelligence
[9] Feng Zeyu, Zhao Erhu. A survey of high school students’ awareness
based on machine learning isinvolved in game development, and needs of artificial intelligence. Electronic World, vol. 552, no. 18,
the future game development mode, gameplay and game pp. 34-35, 2018.
experience will eventually be innovated and changed. [10] Ji Zili, Wang Wenhua. Strategic planning for the development of
military applications of artificial intelligence for world military
powers. Military Digest, vol. 473, no. 17, pp. 9-12, 2020.
[11] Li Haoyuan. Talking about key technologies and applications of game
artificial intelligence. Digital World, vol. 151, no. 5, pp. 451-451,
REFERENCES 2018.
[1] Cai Xinzhang. Wisdom of life across life issues: Life and death [12] Wang Yuan, Li Lei, Dong Haixiang. Research on Application of
education in National Sports University. Student Affairs-Theory and Artificial Neural Network in Games. Electronic World, vol. 589, no. 7,
pp. 11-12, 2020.
[13] Li Tian, Zhang Shumei, Zhao Junli. Design and implementation of
intelligent flow field pathfinding algorithm for real-time strategy
games. Computer Applications, vol. 40, no. 2, pp. 602-607, 2020.