OpenRA: AI Development Guidelines (2018)

github.com

The AI shouldn't cheat, use only information that is available to the player.

I believe that should mean only the audio and video output of the game, just like real humans have to use.

Or implement similar constraints.

I've sometimes wondered what an FPS bot would be like, if it had random noise injected into its aim so that it couldn't just aimbot you; or if it tended to forget where people had been. It could still perfectly strategize about its own limits, but those limits would be unavoidable.

Or an RTS AI with a hard max on APM and some kind of attention span switching delay (like scrolling the map to get to where the action is).

TA Spring, the open source game inspired by Total Annihilation had multiple AI:s, some of them open source. People just developed them as libraries and talked about them on the forum. You could just download and plug them in to the game. You could even pit them against each other.

The current title, "AI Development Guidelines", is too broad.

... summons moderator with some unknown spell ... Would it be possible to rename this submission to something along the lines of "Game AI guidelines for OpenRA (open source Command and Conquer)"? Thanks.

Reply on news.ycombinator.com