Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
0
1
00:00:00,060 --> 00:00:05,990
Before we can discuss decision systems, we must first look at what makes an AI work.
1
2
00:00:06,000 --> 00:00:07,410
What are its parts?
2
3
00:00:08,160 --> 00:00:16,110
An AI Agent is possible because of the following parts: the brain which is responsible for the decision
3
4
00:00:16,110 --> 00:00:16,830
making.
4
5
00:00:16,890 --> 00:00:23,910
The senses or sensors are the AI's way of detecting enemies, obstacles and more.
5
6
00:00:24,240 --> 00:00:28,920
They are linked to the brain to provide the accurate information in real time.
6
7
00:00:28,920 --> 00:00:36,690
The body receives the actions from the brain and executes them, and the legs are mostly used to move
7
8
00:00:36,690 --> 00:00:39,900
the character using different pathfinding algorithms.
8
9
00:00:39,930 --> 00:00:46,650
Let's now focus on the brain of the AI, the decision systems. And there are a couple of them.
9
10
00:00:46,650 --> 00:00:49,290
We will take each one and discuss it a little.
10
11
00:00:49,290 --> 00:00:51,330
Finite State Machines.
11
12
00:00:51,330 --> 00:00:53,310
This AI is used a lot.
12
13
00:00:53,310 --> 00:00:59,790
Actions are mapped to states and the AI can be in a particular state at the specific time.
13
14
00:01:00,210 --> 00:01:07,320
The states are interconnected together with transitions that get triggered when a particular change
14
15
00:01:07,320 --> 00:01:08,700
happens to the agent.
15
16
00:01:08,790 --> 00:01:14,490
Specific games to this are Doom and Quake, but there are a lot of other games using finite state machines
16
17
00:01:15,210 --> 00:01:15,690
Behavior trees
17
18
00:01:15,690 --> 00:01:22,860
Are very common in most game engines, having even a direct integration in Unreal Engine.
18
19
00:01:22,860 --> 00:01:29,160
Their main selling point is that the designer can create any AI system without having to interact with
19
20
00:01:29,160 --> 00:01:30,240
any piece of code.
20
21
00:01:30,330 --> 00:01:32,160
They are very flexible to use.
21
22
00:01:32,610 --> 00:01:36,960
One example for behavior trees is of course Halo.
22
23
00:01:36,990 --> 00:01:39,870
Planning. This one is very similar to state machines.
23
24
00:01:40,050 --> 00:01:46,080
The main difference is that the states that represent the actions are not connected to each other,
24
25
00:01:46,170 --> 00:01:54,270
but rather these are connected real time through a specific algorithm that gets the best possible action
25
26
00:01:54,270 --> 00:01:54,870
sequence.
26
27
00:01:54,930 --> 00:02:01,680
This is not a common way to do AI, but it's found in games like F.E.A.R. Utility AI
27
28
00:02:01,800 --> 00:02:07,230
The utility AI does not have a particular structure but rather a set of rules.
28
29
00:02:07,590 --> 00:02:14,100
It's mostly used to handle complex actions like an RPG character that has multiple skills and needs
29
30
00:02:14,100 --> 00:02:17,160
to pick one or another based on the current situation.
30
31
00:02:17,340 --> 00:02:24,150
This is mostly found in RPG games or card games or games that require the AI to decide from a big list
31
32
00:02:24,150 --> 00:02:24,990
of actions.
32
33
00:02:25,440 --> 00:02:28,320
Games for this are Dragon Age Inquisition.
3454
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.