AI has been constantly evolving since the dawn of computers. Ways of making the computer perform actions by itself, improve learning and eventually, reach intelligence through pervasive AI (such as IBM Watson) have all occurred in the last 70 years. Over the last 40 years, that application has been made available to video games, starting out in the likes of Pong, one of the most basic and primitive of AI, to the complex rulers in Sid Meier’s Civilization series. The needs of an AI system vary greatly, and their inclusion may not even be needed at all.
The purpose of AI in Video Games
From the early days of video games, systems were primitive and handled data that, at maximum, usually only held kilobytes. There was no online multiplayer for decades and the only way to go to create a challenge was to create a situation playable by one person (or two, if the game supported additional local players). Therefore, introducing AI in video games was almost a necessity. As technology improved, AI could be used for other elements of a game, such as shopkeepers or friendly NPCs, in the likes of team members in long running sports games or companions from Final Fantasy. Now, they are used for almost any character you meet in single player games.
In games like Pac-Man, the enemy AI would look for the player and try to hunt them down, or kill them, with very little nuance of unpredictability due to hardware limitations. With Super Mario Bros.(NES), enemies have different movement and attack types, such as moving back and forth or flying through the air. In Starcraft, the AI builds up its base and forces to a point where it believe it can attack the player and will occasionally scout opponents out in order to achieve victory. In Guitar Hero, the AI will hit notes in the song as they appear with varying degrees of accuracy depending on the difficulty. In essence, AI is intended to create a living environment that feels as though it is naturally adapting to the player and the conditions given to it.
Other posts on this blog will relate to how the AI in a game make the game more challenging or believable, likely with video to support this.
AI in an Arena Game
Since their popularity in the 90s, arena games have been a popular place to include many bots to test against before going online and playing against real players. In a fast paced, enclosed environment, it is typical to find 16 people/players against each other in an explosive gib fest. With many difficulties tied to the AI included in games of this genre, like Quake and Unreal Tournament, AI made those players who couldn’t use online functionality get a chance at the frenetic combat of multiplayer. As online capabilities became more prevalent in gaming however, this function started to decline however, one of the more notable releases to include bots was the 2005 Xbox 360 game, Project Dark Zero. The Call of Duty series has also started to include a bot version of it’s multiplayer mode, though only in the last 3/4 iterations so there is definitely an audience for creating AI populated multiplayer-like modes.
Below is footage from Perfect Dark Zero, showing how the bots in multiplayer, at their highest difficulty perform.
While not necessarily fair, these bots provide a high level of difficulty to challenge the player, should they be able to perfect their skills. A step up from the difficulty displayed here is evident in the bots from Counter Strike: Global Offensive as these can auto lock onto your head the moment you appear in their view.
So how does this relate to implementing them into my arena game? They make for an easier way to play around the rhythm mechanics, where they can be set to follow beats within a song effectively, giving a rhythm to both the music and gameplay as a whole. Implementing gameplay that uses players alone makes for a awkward approach to learning, flow can easily be broken by other players not utilising the music at hand. By implementing a single player alternative, using AI systems to create the foundation for a variety of different enemies, it eases players in at first and can become gradually harder with difficulty and enemy types as well as differently paced music. They can come either in wave form or in a standard setup for an arena deathmatch game, such as a 16 player free for all or as two teams.
Aaron Swiss-Hamlett 