Milestone rolled out a new video for the upcoming MotoGP 19. It focuses entirely on the A.N.N.A., middleware suite developed by Orobix, an engineering outfit based out of Bergamo, Italy. Orobix has worked on creating advanced artificial neural network systems, and MotoGP 19 will be the first racing game to make use of their technology.
A.N.N.A., stands for Artificial Neural Network Agent, and the technology is being used for the AI opponents in MotoGP.
The core of the system is based around reinforcement learning of the AI, so that the artificial intelligence learns and evolves to challenge players in a more organic way. You can see how it’s all being implemented with the video developer documentary below.
In the developer documentary, they explain that the AI system is based around a reward system, so that the AI attempts to maintain reward-based performance.
They gave the AI a visual cone in order to identify where it’s positioned on the track and how to dynamically react to other players and opponents on the track.
According to the video, the track times for the A.N.N.A., system rendered results that were several seconds faster than traditional AI opponents. They also make micro-adjustments throughout the game in order to further adapt to the track, the bikes, the player, and the racing conditions. They’ve been testing and re-testing memorization and optimization for the AI in order to achieve peak performance.
Personally, I didn’t see enough to say whether or not it’s anywhere near as advanced as they’re hyping it up to be in the video documentary. The first thing that came to mind was the old Drivatar system that Turn 10 Studios introduced way back in 2006 with Forza Motorsport on the OG Xbox, which operated under a very similar infrastructure, and was further advanced in subsequent entries in the Forza Motorsport franchise.