rec.autos.simulators

At times its very odd to see this, I don't notice this with the N4 AI,
at  The Glen and SearsPoint they appear as I do.

Can anyone explain why the GPL AI appear like this, as though an
invisible hand is guiding them around the track.

thanx,

Rich#8

Democracy is two wolves and a lamb voting on what to have for lunch.
Liberty is a well-armed lamb contesting the vote.

--Benjamin Franklin 1759

Al Unser Jr. said it best about driving the heavy, bulky Winston Cup
cars on a road course: "they don't turn, they don't go and they don't
stop."

1991 -- Alan Kulwicki--On Saturday night racing as opposed to Sunday
afternoons: "Its basically the same,  just darker."

It's because you're using the Glide drivers instead of D3D or Rendition...

OK, I'm going, I'm going :)

--

Tony Whitley
GPLRank -0.54
"On some laps I'd get as many as fif*** corners really right, but Moss, he
gets them all right on every lap" - Von Trips after the 1961 German Grand
Prix.

Booooo!!!
Seriously: The AI is not using the same physics as the player's car
does (or you would need a pc 10 times as fast as your current pc).
So yes, they can do things you can't do. Watch them when they spin,
they can turn their car 180 degrees while barely moving..

Andre

As for why you notice this more in GPL, I guess it's because of the
exagerated driving style of those cars, ie - low grip, high slip angles etc.
Although the GPL AI is as good as you'll see, the cars are on invisible
rails to an extent.  A good spot to see this is the Monaco chicane.

thanx again,

Rich#8

PS:  in porsche unleased, Viper Racing and NASCAR 4 this rail riding
is less pronouced almost non existance.  Anyone know if in WSC are the
AI cars using the same physics as the player?

On Tue, 29 May 2001 15:36:37 GMT, "Simon Brown"

Andre

This is of course true, but I think the reason why the AI are still quite
convincing is that essentially they are following a predefined lap (defined
by a human driving the course), with some variations for car avoidance,
deliberate mistakes etc.

So, they go at more or less the correct relative speed at each point on the
track.

However, sometimes the "base lap" is slow at certain points (eg. Silverstone
at Maggots (T2)), so the AI are artificially slow at that point, which makes
overtaking them a bit predictable!

This is where I find Ratbag's games rather admirable. All their AI drivers
(right from Powerslide) have been created using neural network training. The
lines that their AI drivers take on the highest level should be rather near
optimal.

I think the AI drivers on lower levels are basically the results of neural
networks that have received less training than those on higher levels.

Maybe this is one of the reasons why a track editor hasn't been released
with Leadfoot.

I think this is definitely the kind of direction that games of the future
should take. There should be no need for a human to program in the ideal
line round a track. Firstly, this is rather tedious, and secondly, the
programmed line may be far from optimal. Maybe neural network learning will
be able to take place in real-time, eventually.

Ryan