FORZA MOTORSPORT 5
Turn 10 Studios
Game Designer
(Xbox One | Racing Sim)
November 2013
Performance Index System Tuning
AI Spline Tuning
SYSTEMS DESIGN
Performance Index System Tuning
"Forza uses a classification and 'Performance Index' scoring system to fairly balance cars across the game. This PI system refers to a ranking number based on a car’s best lap time around a simulated test track. The slower the lap time, the lower the PI number; the faster the lap time the higher the PI. This test track is mathematically generated to be representative of the all race tracks in a specific game. We call this a 'PI test track.'" -- ForzaMotorsport.net
Responsibilities
I was tasked with generating and balancing the PI Test Track. This was achieved by gathering and analyzing all Forza 5 track data within MATLAB, and designing a Virtual Test Track that was representative of all the game’s tracks. This data was then used to determine a car's rank within the game, dictating which race events it could qualify for. The functionality could be compared to assigning weight classes to the cars of Forza.
"It's kind of comforting to see that your AI opponents aren't computer perfect. An AI-controlled driver might slide out and hit the dirt in that one sharp turn in Laguna Seca, for example, as that's what its real-life counterpart might have done there. And when you're being beaten, where previous series games might have had you cursing the game AI in general, Forza 5 will have you cursing a specific Drivatar's name. This all completely changes how the game feels as a single-player experience, making it the most engaging racer campaign I've ever played."
Destructoid
AI DESIGN
AI Spline Tuning
"Whereas other racing games use a standard and time-tested form of artificial intelligence, Forza actually learns from the driving behavior of actual players. The data it learns from watching you play is saved as a driving avatar - a Drivatar.” -- ForzaMotorsport.net
Responsibilities
I was charged with generating and balancing AI racing splines on all tracks, at multiple difficulty levels. Being a racing simulation, I utilized real-world data to compare against the AI's race results, with the goal to get as close as possible to real-world lap times with a specific car on a given track. In this case, the Corvette C7 was selected as the test vehicle. In addition to controlling the AI spline positions, I adjusted braking and acceleration zones within the levels for more granular control over the race AI.
In addition to the standard AI race splines, the newly-added Drivatar system utilized custom AI splines to better mimic the behavior Drivatar personas would typically exhibit for each turn on each track. For example, I created a variety of custom splines that would cut a corner too close or take it too wide, based on Drivatars' tendencies.
Video credit
"Thanks to these AI profiles, just about every race pits you against racers that feel human. It was eye-opening being thrown against friends who were playing the game, and finally getting a view into their driving habits."