Original subtitles

0 How does the AI navigate in the game world? 1

The agent is not only just a collection of decision making algorithms. 2

Those decisions need to be applied to specific contexts. 3

For example, the AI needs to move around, avoid obstacles, get to certain targets and things like 4

this. 5

To achieve this, the AI needs to be able to understand the environment where it can move or cannot. 6

What are the points of interest like cover points or important locations? 7

Where is the enemy? 8

While the player can perceive the environment by visual cues like 3D objects in forms of paths, obstacles, 9

the agent needs to get a processed version of that in a language that it understands. 10

If used the exact environment as is, the AI will have a hard time computing everything because usually 11

the environments are too complex. Even in 2D games 12

the complexity of the visual world is too much to handle for an AI instance. 13

I'm not saying that the AI cannot handle it. 14

Most likely it can. 15

But when we are discussing AI Agents, we are always take into consideration the amount of CPU used when 16

running 17

20 agents, 50 agents, or a higher number of agents. 18

We need to get the simplified version of this world. 19

Here are the most common ways to achieve this. 20

Grids: a set of rows and columns that form a 2D table. 21

Graphs: a collection of points that are connected between them. And Navmeshes: 22

A simplified environment, mostly auto generated by modern game engines. 23

Let's look at these three and see their pros and cons.