Pathetic AI: Formula GT At Nurburgring GP

  • Thread starter Thread starter Troux
  • 37 comments
  • 5,354 views
Anyone else notice the difference from the A spec AI and B spec AI?

In b spec if my bob is running at half pace he is fine, anymore pushing from me he decides he is gonna take a stroll in the grass. However in b spec, his fellow drivers almost never pull the same feat is in a spec. Last race I did I think maybe one driver went in the grass and spun out because my bob forced him into the grass.

A-Spec AI is somewhat faster than them in B-Spec, hence this chicane problem magnified...:dunce:
 
I think the most interesting part in the FGT races is the boosting. I mean notice how the same couple of cars with be right behind you no matter how many times they spin out.

There is no boosting of AI, or my game doesn't have it somehow.

My Bob is level 20 going on 21, and in every race I've watched, he beat the AI by an average of 17 seconds. If you leave your driver to full blue, the AI will eventually catch up because they are either battling other cars and get into the red zone, or the them being higher levels allows them to drive faster at full blue then your driver.

Bob is kept at neutral or a smidgen on the blue or red side, and the AI, after a couple of laps, stop trying to catch up to me. My latest race in Monaco saw me win by 34 sec.
 
They did not stop to throw me out road in the Schumacher S curve, it was quite annoying.
 
So... GT5 has the best and worst AI of any racing game... AT THE SAME TIME?

Whoa.
 
At least it is realistic in that kinda way that AI doesn't drive too robotically.
I mean that they even crash sometimes. In many games the AI never crashes which is boring and unrealistic.
Even Sebastian Vettel crashed, although he is the best drivers in the world.
 
I've noticed if you do a race at Deep Forest in the Ferarri F10, about half of the AI's crash coming out of Turn 9.


And then whatever cars don't crash, and any cars that manage to get turned around, all crash at turn 14.
 
Same with the new Seasonal in Cote d' Azur. It's so funny to see 2 AI Lotuses spin out every lap xD
 
I believe the AI you describe is the kind of AI most games have. Basically, you set rules, which are track and car dependent. You work an equation to scale them so they work fairly well with similar cars. But this is where the problems start to show up. You'll probably need different rules for different types of cars since your scaling equation will probably not work with them, maybe you need different equations too, maybe depending on the kind of drivetrain you'll need different rules, etc... The rules are basically the "telling the AI what to do".

In a game with 1000 cars from very different categories and quite a few tracks, this can easily be blown out of proportion. They would waste a lot of time making those rules. They need a different way since this one is probably not feasible. And what about the automatically generated tracks and rally stages? The AI wouldn't be able to race those since there would be no rules for them.

What I believe they did is what Griffith500 described. Basically, you try to teach the AI how to drive. This way, it will work everywhere, even at the automatically generated courses. However, making an AI that can drive isn't easy. With so many different cars and tracks it's no wonder it will struggle at some point.

But again, this is just my theory.

There is no way PD will have created such a complex AI system that it "learns" the track.

The fact there are several bugged areas of tracks proves this, if it "learned", surely it would learn not to keep spinning or crashing in the same place?

Quite clearly, the AI runs a pre-determined route with adjustments made for braking and turning points depending on the car (or rather, based on the speed of the car and some other variables). Considering the game has a dynamic racing line assist that adapts braking points, it wouldn't be unreasonable to assume that AI also use this.
But the line the AI takes will almost always be the same exact line every single lap. Effectively the AI programming will be "follow the racing line" with conditions to cope with overtaking and avoiding other cars.

Its interesting if you ride onboard with B-spec and observe where the AI brake and accelerate. You will notice on a fair few corners they will not accelerate even if the car they are in can easily make the corner until they are past a certain point.

While the AI does make the odd error and does stray from the racing line from time to time, this can be down to many factors. But in ideal conditions (full strength, fresh tyres, no cars to overtake, etc) the AI will follow the racing line perfectly, everytime.

The Nurburgring bug (And the Monaco bug) will almost certainly be down to how the racing line is generated for those tracks, the throttle point for the chicane at the Ring is clearly too early for high-end cars like the FGT. There might also be a bug with how the AI follows the racing line in this particular section, it might be the speed it takes with the FGT confuses it and it loses track of where it is (perhaps there is a bug where the racing line momentarily disappears for the AI and its left to run off the track before the racing line appears for it again and it can correct).

As for the "randomly generated" circuits, you can have pre-programmed AI racing lines for these tracks. If we assume the AI runs off the dynamic racing line, then it will simply follow the dynamic racing generated for these tracks. Why bugs don't appear on these tracks so often (though they do appear) is probably due to their being a pre-set list of corners that the Track Maker can use to create the circuit. But I think its very likely that these tracks will have some kind of automatic racing line generated for the AI to follow. If the Track Maker can only produce so many different types of corner, its possible to produce a generic racing line with variables for elevation, degrees, etc.

No need for excuses, I think its fair to assume PD will have come across similar bugs for other tracks and will have fixed them.

There are very few if any video games which use truly complex AI models, its so time-consuming and incredibly unreliable, unpredictable and difficult to perfect that its not sensible to try to make "intelligent" AI. The kinds of knowledge systems we are talking about are huge! (1000s of cars, thousands of corners with billions of different scenarios!!)
AI in most if not all games is an illusion, nothing more.
 
Last edited:
Back