Detroit downtown circuit (WIP thread)

[checkpoint10]

Have looked around in my files. I was certain I had a Detroit sg file somewhere, but couldn't find it. Might be because I cleaned up a bit over time.

This one I could find in my GPL files. It is the GTK file from GPL, which is the result of the conversion from GP2/3 to GPL. I would convert that after some fixing to a GPL trk file. From there you can use the trk2sg utility that is on the forum/archive here to convert from GPL to ICR2.

The layout of this GP2/3 conversion looks inaccurate in many places though. Compare to the map on Wikipedia: https://en.wikipedia.org/wiki/Detroit_street_circuit (even though that map itself is also slightly inaccurate in its portrayal of Turn 1 as two 90 degree turns when it is one turn, and also I believe it has got the Turns 3 and 4 section wrong, but at least everything else is overlayed onto a street map).

[samsepi0l]

Turn 1 is a long constant radius corner, taken in probably 3rd gear. I agree with Checkpoint10 here, as it looks like 2 right angle corners you would see in houston in the map here. The gp2 version of the track got this right.

[Dennis]

Two other puppies I found

I know the second one. But what is the first one? Should I know it or is it not so famous?

[checkpoint10]

Two other puppies I found

I know the second one. But what is the first one? Should I know it or is it not so famous?

Sebring and Le Mans. Both would be nice to have in ICR2 eventually!

---

Turn 1 is a long constant radius corner, taken in probably 3rd gear. I agree with Checkpoint10 here, as it looks like 2 right angle corners you would see in houston in the map here. The gp2 version of the track got this right.

I will stitch together the aerial photos (probably from 12/2002 on Google Earth) and use that to draw the track and walls on another layer. This may need to be a fully scratch build.

I also drove Stig's version a few times and opened in SGE and I think the corner radii are a bit large making the turns longer and the straights shorter than they should be. The track length is correct though. I know that sharp turns in ICR2 are challenging to do and may cause AI issues, so hopefully we can still make the turns as sharp as they are in real life.

I wonder if we should focus on getting track geometry right, and then do the AI right afterwards, in order to verify that we don't need to modify the geometry, before proceeding with trackside objects.

[Dennis]

I wonder if we should focus on getting track geometry right, and then do the AI right afterwards, in order to verify that we don't need to modify the geometry, before proceeding with trackside objects.

I think, it isn't necessary to make the final .lp files before making track improvements. It's enough to make some fast test laps after the basic geometry is complete, to look for high Coriolis values or incomplete .txt files made by trafo20, and then make the necessary geometry corrections. As I made my track, I saw already after some laps, that the .txt files are wrong.

[checkpoint10]

I think, it isn't necessary to make the final .lp files before making track improvements. It's enough to make some fast test laps after the basic geometry is complete, to look for high Coriolis values or incomplete .txt files made by trafo20, and then make the necessary geometry corrections. As I made my track, I saw already after some laps, that the .txt files are wrong.

Dennis, in your opinion, do you see any potential problem spots just from looking at the circuit map? https://en.wikipedia.org/wiki/Detroit_street_circuit

Since we're doing the IndyCar layout, it will ignore the final chicane before the pits, otherwise it's the same as the F1 layout.

[Dennis]

It depends on the track width. If the track is broad, all 90 degrees corners have a problem potential, because of higher speed at small corner radius.

[samsepi0l]

Turn 1 goes through a parking lot. On most of the maps it shows as 2 90 degree turns, but it's not....

Guys- watch this lap with Danny Sullivan- you can clearly see it is all one long sweeping corner:
https://youtu.be/NZgCf_yIscc?t=411

Check this preview out from 1990 with Johnny Rutherford.... (super detail in this corner)
https://youtu.be/hc8NG4K8dxQ?t=395

Dennis- with your formula method, I usually have no problems with AI in the tight corners because I'm able to correct the speeds based on the corner radius and position.

As far as the many 90 degree corners that we have to do-- I suggest looking at the radius of similar corners on factory icr2 tracks. One that I'm very familiar would be in toronto, after the hairpin at lake shore blvd. (where they turn left 90 degrees) There is also a good 90 degree corner at the end of the back straight in Toronto we can see how the original track was made, and understand good radii to use. (I actually have home footage from being at this race in 1994 standing at this corner in real life)

Maybe I am a complete naive fool- but I think I can make the AI work with any geometry using my spreadsheet method combined with Dennis's formulas.

---

best map I can find:

[Dennis]

Tight corners are no problem on their own, they come to a problem paired with high speed, when you switch position fast from one side of the track to the another (broad track). Then you get high Coriolis values and the .txt file which you get from your replay has from that point on no relevant features. If you haven't see this effect till now, maybe you haven't such tight corners.

A solution could be to rise the corner radius, make two sections for the corner, and put the track surface at the apex more to the outside, so the position data in the replay don't change so much. You just need to test it, how such a corner look like in the game.

[checkpoint10]

@samsepi0l The course map is really helpful - thanks for finding it! The only part that I'm not sure about on the map is the right turn onto Woodbridge. Based on video, the course actually goes onto Woodbridge proper (instead of along the edge as per the map), and then makes a crossover onto Jefferson Ave. However, there seems to be no good video I've found of that transition to Jefferson.