Space colony sim MMO SEED offers a look at how its AI operates

    
2

Alessia Nigretti, one of the AI engineers of the in-development space colonization MMO SEED, seems pretty happy about her job and eager to share her work. In a thread on Twitter, she offers some insight into how the game’s Utility AI system steers the characters in the game.

The AI logic of SEED can be broken down into three parts: Discovery, Selection, and Execution. During Discovery, items in the game world are effectively asking to be interacted with, but filters need to be applied via Selection, which takes into account whether a task is impossible (ie. You can’t eat an apple that someone else is eating). Even after that filtering, there are still a lot of tasks left, so a scoring system is then factored into each task, with the highest scoring one being Executed. Each activity isn’t scripted, but is made up of simple atomic steps that the designers of SEED can pick and choose to create new behaviors to simply let colonists flourish.

“I didn’t know about Utility AI before joining my team, but now I’m a huge fan!” Nigretti writes. “Utility is the perfect type of AI for games that are expected to be highly scalable and data-heavy, and that plan to put some focus on emerging behaviors in rich and complex environments.”

The whole thread is worth a read for armchair devs as well as to bask in the pure happy nerd energy, but what also might be interesting for MMO fans is that Nigretti gives thanks to talks from AI developer Dave Mark, who many readers might remember was one of the AI authors of EverQuest Next. Mark was struck by a car at GDC 2018 and nearly killed, but is now working as a game AI consultant. As for SEED itself, the game is expected to kick off alpha testing at some point this year.

source: Twitter
Advertisement
Previous articleMad World interviews its writing team and further fleshes out its universe
Next articleStar Stable has added new quests, updated areas, and introduced new horses

No posts to display

2 Comments
newest
oldest most liked
Inline Feedback
View all comments