Game & AI Analysis

Learning:

Artificial Intelligence: in gaming terms can be understood as getting a machine to perform tasks in an intelligent way; however, when considering player mental models of games - good game AI, can be said to be getting a machine to perform tasks that suite the perceptions and mental model, a player has, of how gaming artefacts should behave. For example, a human will be expected to be able to stand, walk, run, jump, open doors, search, wander, talk and not see through walls, unless using another gaming artefact; and a games AI must do all these things, while appropriately switching between them to suite the situation.

Core capabilities of AI:

1. Perception, information matches simulated entity, for example a human cannot fly or see through walls. However, a Non-Playable Character(NPC) can also have a perception created by the data the entity has access to from a games data base.

2. Think, a simulated entity must evaluate it information, and plan according to goals it has.

3. Act, a planned action if exercise often and through the use of animations, for instance attacking the player when seen, still requires movement to the player, which can be a run or martial footwork depending on distance, and attacking itself can have a host of animation, again based on information, or even random number (when considering the complexity fallacy below).

Complexity Fallacy: more complexity is not necessarily best, if the player is not interacting with it; for instance creating a driving game with more complex driving skills such as turning and braking would be appreciated, however, if the complexity come about via a background economic system that the player does not interact with at an intrinsic level, one would not be able to tell the difference between the use of the complex system or a random number generator. Resulting in the complexity being a waste of resources, and imperceptible to the player mental model of the game.

AI techniques:

There are a range of AI techniques that can be used in games, for example Gamasutra (2018) in the work [7 examples of game AI that every developer should study] considers a range AI techniques and its suitability in games, and would agree with the pathfinding and neural network commentary below.

Pathfinding: Division's AI-driven path finding – not merely for moving from point A to B, but how it can be used as a game mechanic to increase player immersion; being used as a method to allow the player to move to cover spot and free up players cognitive for other aspect of the game such as developing strategies to deal with enemies.

Neural Networks: Forza's Drivatar adaptive AI system - allows NPC drivers abilities to be defined by players actions in the game environment and create unique NPC’s who intern create unique challenges for players. Essentially create a NPC driver based on the Players from around the world.

Decision Making: is distinctly how a computer or program will make decisions in games; there are a range of AI techniques that can used for instance Mount, D (2013) discusses the transition from Finite State Machines (FSM) to Hierarchal FSM (HFSM); and other additional approaches such as Behaviour Trees and more.

Presentation:

To view presentation, please click on link - PDF

References:

Gamasutra (2018) 7 examples of game AI that every developer should study [online] available at <https://www.gamasutra.com/view/news/269634/7_examples_of_game_AI_that_every_developer_should_study.php>[25/01/2018]

Mount, D (2013) CMSC 425: Lecture 20 Artificial Intelligence for Games: Decision Making [online] available at <https://www.cs.umd.edu/class/spring2013/cmsc425/Lects/lect20.pdf>[20/01/2018]

Bibliography:

Finite State Machine and Decision Making [online] available at <http://slideplayer.com/slide/1544892/>[26/01/2018]

#GamesandAI