CORIOLIS

[CowtownBob]

Just a thought...

No idea how closely Papy followed it or even if they used it and how, but keep in mind that Coriolis isn't just a word; the Coriolis Effect is a real-world phenomenon and quantifiable by formula.

It's an effect whereby a mass moving in a rotating system experiences a force (the Coriolis force) acting perpendicular to the direction of motion and to the axis of rotation. Or, what's commonly known as lateral G force, if I read it correctly.

Fc = -2m(Ω × V)
Where Fc is the Coriolis force, m is mass, Ω is angular velocity (how fast is the car turning), V is tangential velocity (straight-line speed)

So maybe all y'all need to do is figure out what's the mass of a pixel indycar.

[Dennis]

I know, what the Coriolis effect is, I had it in my study. It is, as you wrote, in a rotating system. But we have here non-zero values also on the straights.

[samsepi0l]

I think, if you make the value zero or too low, that happens, what we can see in your video.
One thing, for that the Coriolis value could be used is, calculating the angle of the car to the centerline. If it is so, with zero Coriolis value but changing position value, the cars would move sideways, instead pointing to the driving direction. But I think, it's not the only thing, because calculating the car angle could be made probably easier.

The reason, why cars are move away from the racing line is another car in the way. But, after they moved to the passing line, they stay there for longer as necessary, and I don't know, on what depends, when they are changing back to the racing line. Is it, when there is another car in the way, or is it, where the passing line is close enough to the racing line, or where the racing line is faster, or when they are far enough in front of the overtaken car?

Your suggestion of coriolis providing "angle" is very interesting... it should be easy to prove with testing. I'm surprised I never thought of this before. I wonder if this is why some of my LPs tend to have the cars "dart" a little bit after they hit the apex of a corner. I am thinking of the LPs I made for caesars for example.

As for your explanation of the long beach hairpin movement- I think you are right in some cases, but I have also seen examples where CC's will not move to RACE.LP when we would think that they should. I did extensive testing in TJERKS glenboot track with a situation like this. To do my testing, I reduced the number of opponents to only 1, and I let them drive around the track all by themselves.

In this experiment there was one specific corner where the CC would always miss the apex of a corner, and run wide here. In all of the other corners the CC drove as I wanted them to. I found that this was being caused by MAXRACE being too close to RACE in the corner previous. When I made MAXRACE a little closer to MAXPANIC it fixed it.

All of my testing has taught me that sometimes when the cars attempt to drive on the LPs as you've given them, if the speed or racing line is too aggressive it's almost like sometimes they will "slip off" of the LP and end up on another LP file. If you really screw this up, they will bounce off of MAXPANIC (or MINPANIC, if were talking about the right side of the track) like hitting a rubber-band. You can see this in the video that checkpoint10 posted where the CCs violently hit the wall.

I think some of the behavior in long beach can be explained by the cars "slipping off" of the race.lp and ending up on the maxrace.lp. We can know for sure with 100% certainty this is (or is not) the case by reducing the number of opponent cars to only 1, and then letting them drive and watching the replays. I like to use the "hold ALT and F12" trick to make time fly in dosbox so I can get 10 minutes of replay in a matter of seconds.

How I like to proceed, is to first make a RACE.LP. Once I have a RACE.LP I want to try, I copy the RACE.LP into my track to test it 3 times. (so I have the same file in there 3 times). I temporarily move or rename the old MAXRACE and MINRACE and replace those with my new RACE.LP- this way I can see for sure how good the new race.lp is because the CCs will have to use it (since in reality it is the same file, just with 3 different names). Once I am happy with the RACE.LP, I can go back and make a good MAXRACE and MINRACE. If something goes wrong then I know MAXRACE and MINRACE need to be modified (or made wider).

Testing with only 1 car really helps.

[Dennis]

All of my testing has taught me that sometimes when the cars attempt to drive on the LPs as you've given them, if the speed or racing line is too aggressive it's almost like sometimes they will "slip off" of the LP and end up on another LP file.  If you really screw this up, they will bounce off of MAXPANIC (or MINPANIC, if were talking about the right side of the track) like hitting a rubber-band.  You can see this in the video that checkpoint10 posted where the CCs violently hit the wall.

Yes. I made a little experiment with rainy conditions this week, but I didn't look at the replay. Now I did it, and there's the same behavior. But in such extreme cases, they try to come back as soon as possible, and don't stay on the wrong line.

Your theory with MINRACE and MAXRACE sounds really interesting.

[samsepi0l]

All of my testing has taught me that sometimes when the cars attempt to drive on the LPs as you've given them, if the speed or racing line is too aggressive it's almost like sometimes they will "slip off" of the LP and end up on another LP file.  If you really screw this up, they will bounce off of MAXPANIC (or MINPANIC, if were talking about the right side of the track) like hitting a rubber-band.  You can see this in the video that checkpoint10 posted where the CCs violently hit the wall.

Yes. I made a little experiment with rainy conditions this week, but I didn't look at the replay. Now I did it, and there's the same behavior. But in such extreme cases, they try to come back as soon as possible, and don't stay on the wrong line.

Your theory with MINRACE and MAXRACE sounds really interesting.

With all of the things we've learned about AI, it is really obvious how nice life would be if we could "see" which LP the CC was using. 

In a fantasy world I would love a tool that puts a little text-box above the car in the replay that says "RACE" or "MAXRACE" or such..


I have learned that sometimes we make LPs, and then when we try them in the game and immediately we see a problem and think "darn- my race.lp is no good, I must make changes"

In reality it may not be a problem with your race.lp, it may be that the cars are moving from race.lp to minrace or maxrace because of some unknown reason. Changing the race.lp is not always the answer. This is where making LPs is frustrating and it turns people off to it.