With the release of last year’s Transformers: Fall of Cybertron, the game was more than just a critical hit. It set a new benchmark for innovation in game artificial intelligence.
As a programmer at developer High Moon Studios in Carlsbad, California, DigiPen graduate Mike Kersey (2007) played an instrumental role in both areas of the game’s success. In addition to helping implement the A.I. systems for Fall of Cyberton‘s varied enemy types, he also engineered and helped design the game’s spectacular and critically-acclaimed final level.
In January, Kersey’s team received the 2012 “Technical Innovation in Game A.I.” award from aigamedev.com, a professional resource and online community for game developers working in the field of artificial intelligence. Fall of Cybertronwas recognized for its forward-thinking implementation of hierarchical task networks (HTN), a type of A.I. planning system used to determine how a computerized character will behave at any given moment during gameplay.
“I was always fascinated with artificial intelligence ever since my first semester at DigiPen,” Kersey says. “It’s exciting taking a character from an initial concept to playable A.I. behavior. I enjoy working with animators, designers, and artists to bring characters to life.”
As development for Fall of Cybertron entered the home stretch, Kersey was tasked with engineering what would end up being the game’s grand finale.
“I was paired up with a designer, and we received some high-level direction from the leads,” Kersey says. “We kind of had free rein to co-create this final level experience, but it was under a very strict, short deadline.”
I was always fascinated with artificial intelligence ever since my first semester at DigiPen.”
While the game’s predecessor, Transformers: War for Cybertron, had been well received, one of the common criticisms leveled against it had been a perceived repetition in gameplay mechanics. In designing Fall of Cybertron, the developers at High Moon responded to this feedback by creating a much larger variety of mission and character types for players to experience.
Players could pulverize their way through an enemy stronghold as the lumbering robot titan Bruticus, transform into a metal T-Rex as Grimlock, fly around in an aerial dogfight as Jetfire, and much more.
“We had a variety of mechanics available that players would have mastered from completing the earlier levels,” Kersey says. “We decided with the limited time left that stringing those mechanics into a montage of short sequences would be an effective use of resources.”
In almost rapid-fire succession, players switch characters, fighting alternately as both the heroic Autobots and villainous Decepticons, until the final showdown between the two faction leaders, Optimus Prime and Megatron.
The finale was a huge hit. Several critical outlets, including the New York Times, Gamespot, and Giant Bomb, applauded the final level as a highlight moment. G4 TV wrote, “The thirteen level campaign feels like it’s building to a massive confrontation the whole time, and that’s exactly what happens in the stunning final level.”
So how did the two developers manage to pull off such an incredible sequence in a five-month period?
All the big games — they do this. Prototype the game mechanic, test it, observe and take feedback, and then iterate.”
In part, it was made possible thanks to the innovations that Kersey and his team of fellow A.I. programmers had achieved through their use of HTN — a technique that allowed for faster prototyping of new character behaviors.
Equally important was their reliance on rapid prototyping and playtester feedback.
“All the big games — they do this. Prototype the game mechanic, test it, observe and take feedback, and then iterate,” Kersey says. “And you’ll keep repeating this loop.”
By iterating quickly and testing new changes, Kersey says he was able to identify and improve on each of the areas that proved problematic or confusing for players. Adding a health bar during a particular fight sequence, for example, let players know that a previously invulnerable boss was now open to attack.
It’s a proven method, Kersey says, for creating a highly polished playing experience — a lesson that applies even to student games.
“When observing focus tests, you begin to realize many players might not play the game the way you intended, or break it in a way you never tried before,” he says. “It’s a powerful tool that helps yield a consistent, fun gameplay experience for a variety of player styles and skill levels.”
The curriculum builds a solid engineering foundation from the ground up.”
In some ways, Kersey says, attending DigiPen was a humbling experience. Having already achieved a modest level of success as a game developer (at age 14, one of his shareware games was published in the U.S. and Japan) Kersey was blown away by the influx of knowledge he received in just his first year of the B.S. in Computer Science in Real-Time Interactive Simulation program.
And while it was rarely easy, he emerged with the skills and experience necessary to land an exciting job in a location he had always wanted to live.
“The curriculum builds a solid engineering foundation from the ground up. You get to stretch your imagination on game projects while learning valuable lessons about the development cycle, teamwork, and social dynamics,” Kersey says. “It was DigiPen’s combination of talented faculty, targeted courses, and multiple game projects that helped me learn the specialized knowledge necessary to start my journey as a professional game developer.”