Artificial intelligence (AI) has been a major disrupter in many industries, but one area where it's excelling is in racing video games. Traditionally, racing games had AI opponents stuck to pre-determined racing lines, making the gameplay predictable and dull. However, with the advent of deep learning AI networks, the racing gaming landscape is tossing out the monotony and embracing unpredictability and excitement.
The traditional predictability of AI in racing games
Gone are the days when beating AI opponents in racing games felt like trying to outpace a train on a predestined track. In the past, game developers drew out a racing line, which the AI followed religiously. The AI would apply throttle and brake inputs exactly as instructed, deviating only when the player forced them off the line. This made offline races seem like a procession, with the thrill of competition diluted. However, this status quo is changing as developers begin to utilize deep learning AI networks to make the gaming experience more dynamic.
Using deep learning for smarter AI competition
Today, game developers are leveraging the power of deep learning to make AI drivers smarter and the races more exhilarating. Developers of popular titles like Gran Turismo 7, MotoGP series, and Forza Motorsport are all employing learning AI, leading to an unpredictable racing experience. This approach disrupts the predictable old structure by giving the AI the ability to learn, adapt, and thus provide a more competitive gaming environment. Each race becomes a new challenge, and player victory is no longer an inevitable outcome.
Sony's innovative 'Sophy' AI in Gran Turismo
A notable example of this shift is Sony’s AI research division's work on Gran Turismo’s ‘Sophy’ AI. Instead of sticking to a pre-programmed racing line, Sophy is given a plethora of information about the car and the track. It's incentivised to drive efficiently and professionally, and to complete laps as quickly as possible. Variables like the car's velocity, acceleration, tire load, and slip angles, along with track progress, surface incline, and upcoming course markers, are all taken into account. The result is an AI driver that races not just speedily, but also strategically, just like a human driver would.
Forza Motorsport 8 takes a similar approach to AI training. Unlike previous versions where AI drivers adhered to one of three possible racing lines, the upcoming version has AI drivers that are far more adaptive and versatile. They are given a set of information and objectives, and then set free to find the fastest way around the track. Having driven the tracks tens of thousands of times, they have developed a multitude of different racing lines across each circuit. This should result in more challenging races, as the AI drivers can now adapt to changing conditions like cold tires and rain.