ML vs AI for a Battle Royale Type Game

News & General Discussion
1

Hello Everyone,
im looking for a way to create NPCs for a simple battle royale 2.5D game eg:

The Agents behaviour should be able to:

  • weapons (Collect Weapons when they are at a higher lvl or no more ammo)
  • Team Mates (heal them when they are dying, Help them when they are fighting)
  • run away from danger Zones ( the game is closing the world map over time to regroup clients in little regions, all agents out of the safe zone will die slowly)

does unity offers any solution for these requirements ??

i have a very beginner knowledge about AI Planner and ML-Agents Toolkit and i have no idea which one i should use.
what technologie would you use and why?

Thank you!!!

1 Like
2

Everything you’ve described can be done without that advanced AI learning stuff. Because you won’t get a reliable results out of it.

Personally, I suggest looking into Behaviour Designer.

Half price for the black friday as well.

3 Likes
3

@Opeth001
Behavior trees
or,
Utility AI
had some discussion here if anyone interested ** Utility AI (Discussion) **,
or,
mix

2 Likes
4

We are creating advanced tactical combat AI, we are doing a BT/Utility approach. In simple terms Utility AI is just BT with a bit softer more fuzzy logic

2 Likes
5

all your approach seems good to me.

in my case i want NPCs to maybe select multiple choises simultaneously, like
*decide to heal a team mate (stay near to a dying Friend),
*Heal him self (pickup a heal)
*shoot the enemy

also the ServerSide which controls all AI and send inputs to clients is using Pure ECS,
so which one can fill these requirements ?

Thank you Again!!!

1 Like
6

Thats an implementation detail. Both can do it.

2 Likes
7

What did you end up chosing?