I found AIs are either easily crash into each other without braking applied or braking to a even slower state when following other cars and accelerate far behind out of the corner when drive through low speed corners in D42. In some case it more seems like a logic of line finding problem. Maybe have a try using a wider line to avoid？eg.Turn 4 and Turn 8 in WE1

And also in D42 I feel AIs are driving more natually and I've had some fun competing with some faster cars.

Quote from BorislavB :

I saw that pit limiter light its already updated to flashing.
Is it possible to make when you turn your lights and press flashing button they still to can flash, because whey they are ON now they not flash, only when lights are OFF they can flash.

I can't take take requests on this thread. It is only for reporting issues with the test patch, that were not there before the test patch.

For feature requests please use the special forum section. https://www.lfs.net/forum/8-Improvement-Suggestions

Quote from Scawen :

Thanks, everyone!

Pasci, can you tell me an example of a track where the N.400 GT seems so slow? There is one thing I am trying that is specific to the public version, and that is braking prediction. Late braking followed by wide cornering is spoiling too much AI driving. I wonder if this could be the cause of the N.400 GT problem.

Sorry for the late answer - I'm now at home.

The track was Westhill International. If helpful, I can provide the SPR file (compressed around 9 MB).

Quote from Scawen :

Maybe you could use a separate LFS instance.

Hmm... interesting approach.

Quote from Scawen :

About the speed of AI, it's not your fault but I'm getting a bit tired of explaining so many times, the AI's driving ability (handling of one car, nothing to do with the overtaking) is all specific to the tyre model in question and has already been improved to some extent in the development version, but I will not be working on that for the public version, as it would be pure wasted time.

Sorry, I guess I didn't understand that until now.

Quote from Scawen :

But you could verify by starting an earlier exe, run a single AI and see if the lap times are the same. That's your job as a tester really.

I can't add anything more to that.

Thanks, I'll compare the N.400 GT at Westhill International in my new version and D42.

I'm feeling a bit frustrated right now as the new code has seriously improved some cases but made others a bit worse. Seems like there's never an easy day when working with the AI code.

Just to make sure: are AI driving the N400 GT mod actually super slow, or are they just driving on the default setup (which is using sports tires) while you are using slick tires? I just started a race around Blackwood with the slick tires, and noticed the AI were really slow, until I forced them to use my setup - they're still slow, but definitely not *that* slow on slicks.

Quote from Scawen :

I'm feeling a bit frustrated right now as the new code has seriously improved some cases but made others a bit worse. Seems like there's never an easy day when working with the AI code.

My few tests are too partial to be meaningful, and patch D42 may have generated certain bugs. But from my point of view, this patch is a real revolution for solo races. I have never had so much fun driving with AI in LFS.
Thank you for this great work.

Quote from Scawen :

Seems like there's never an easy day when working with the AI code.

Trying not to over post here but I'm sure I speak for the vast majority when I say that your efforts on this are greatly appreciated.

Also it seems to me that when the new physics and graphics are released there will be a sudden focus on LFS and the first thing that reviewers and YouTube Vloggers will do is jump into an AI race so it will be a downer if they all crash on the first corner. However if it is a better experience it should have a positive affect on your bank balance .

On a practical note, it seems to be a downhill braking problem so could there be a compensation factor for the incline? Also, if a human driver locked a wheel they would release the pressure and reapply the brake whilst perhaps taking a wider line - could this be simulated?

Just some thoughts ...

Thank you all for the observations, reports and comments, even if I don't reply to all of them.

I think I've got some decent results now for the braking, and should be able to release that tomorrow.

Quote from Scawen :

Thanks, I'll compare the N.400 GT at Westhill International in my new version and D42.

I'm the author of the mod, if there's anything I can do to help, let me know.
I'm too in the process of testing AI with my mod right now.

On latest D42 (but observed also on D40, not sure if previus issue exist) when AI enters pitlane, it exceeds speed limit and gets penalty. Observed on Rockingham and Westhill. It happens when AI is changing gear from 1st to 2nd, for split second having too much speed.

Car : Prototype-2

Edit: I've just realised that I've been generating paths on this track for the same cars on the on 07D40, so its a 07D40 test even im on 07D42 version right now. Tomorow will make same test in 07D42 on other instance of LFS.

And I don't feel that the N.400S GT4, behaves significantly worse compared to other mods in the same power class. I've also added an XR GTR just for fun.

N400S GT4 Made according to the regulations of real racing in GT4 class
All white\gray cars on sport tire, all red\black on slicks

Start positions:
AI-1 N.400S GT4 WHITE (SPORT TIRE 327 hp\ton)
AI-29 FZ50 V8 Safetycar (SPORT TIRE 319 hp\ton)
AI-5 FZ50 V8 Safetycar (SLICK TIRE 319 hp\ton)
AI-32 N.400S GT4 (SPORT TIRE 327 hp\ton)
AI-6 XR GTR (SPORT TIRE 453 hp\ton) (make it with sport tire throught alternative config)
AI-2 N.400S GT4 (SLICK TIRE 327 hp\ton)
AI-27 N.S80 (SPORT TIRE 412 hp\ton)
AI-3 N.400S GT4 (SLICK TIRE 327 hp\ton)
AI-7 XR GTR (SLICK TIRE 453 hp\ton)

Results in pic. Replay attached.

I like that overtaking is now more aggressive, but I don't see any significant flaws in my mod specifically relative to others.

UPD: Forgot to mention that the all mods are on the default setups.

Quote from Aleksandr_124rus :

N.400S GT4

The car lacks Traction Control (TC) and that's likely the main reason why the AIs struggle to drive it well.

On D42, if an AI stalls at the start it doesn't notice, it'll still change gears and use throttle as if the engine is running. If it's downhill it'll still steer as it rolls, foot to the floor... bless

The stall can come from a little wave of panic that goes through a grid of large vehicles (or standard-size vehicles on a custom, tight grid). The revs are just at their lowest from launch when everyone other than the front row brake to avoid the one in front. They often get away with it, sometimes just barely.

I don't know if the grid-panic itself is worth any attention given that it can be avoided with gearing and/or grid spacing.

The RACHA 707 will show it up with a default set on the standard start grids of AS1R, FE1, WE1.

Quote from matze54564 :

The car lacks Traction Control (TC) and that's likely the main reason why the AIs struggle to drive it well.

Well in the mod there is traction control, but it is not in the default settings on which the bots in my test. And they managed to drive similarly to other mods.

Two new tests for 07D42. It's also important to mention that all the tests I'm doing on the lvl5 AI on all cars.

Second test - same as the first test, same track, same 3 laps, same cars, same placement (but different AI names), same setups, but there + one (AI 10) N.400S GT4 on Slicks with traction control to see is there any difference.


No special changes, or maybe my mod handles even better. Bot with traction control drive the same as bot with no traction control.

The third test - for testing traction control.

AI 10 N.400S GT4 (Sport)
AI 11 N.400S GT4 (Sport + Traction control)
AI 12 N.400S GT4 (Slicks)
AI 13 N.400S GT4 (Slicks + Traction control)

I see almost no difference between bots with traction control with bots and without. Bots with traction control have no advantage over those without it.

I see already 3 people saying in this thread that something is wrong with my mod. I'm trying to understand what's wrong, but I don't see what's wrong with my mod.

Thanks for the tests. I don't think that people are saying that something is wrong with your mod. Only that the AI had difficulty driving it to its full potential. I think it's because of the high power with road super tyres. It is a well liked mod and a good test case, so that's why it got attention.

I've been using it for some tests today and now, through the braking code adjustments, it is about 0.7 seconds faster at Westhill International.

Racing alone, lap 2 of 4 lap race:
D42: 2:13.51
NEW: 2:12.84

If other tests work out, this is quite a good result. The new braking prediction is designed to handle brake balance better. I got an XRT and set brake balance to max front and max rear and in both cases it could actually take the troublesome left hander at SO Sprint 2 without clouting the wall. teeembo's Lotus inspired F1 car seems to take all corners on its racing line now.

All right, here is D43

The main thing is the improved braking estimation, then today I had a look at the reported issues that seemed most important or easiest to fix.

Thank you again for all the testing and comments.

I hope I haven't added any new problems!

Changes in D43:

Improved braking prediction so less running wide at corners
- considers brake balance (which is not ideal for every corner)
- can result in better lap times due to improved line following

Target speed 1 km/h slower in pit lane to avoid speeding by mistake
- was possible for a powerful car to overspeed shifting 1st to 2nd

Distance to vehicle considered safe is reduced at low speed
- should prevent long vehicles hitting the brakes on green light
- gaps between vehicles may be smaller when speed is below 20 m/s

FIX: Can now reset at the end of a pit stop (e.g. fallen bike)
FIX: It was possible for the fuel calculation to report 0 stints

https://www.lfs.net/forum/thread/102117

Should we recalculate AI paths on newer patches? If yes, how to do it? Which files should be deleted?

No, it shouldn't be necessary.

In fact, due to the very slightly different braking calculation, the optimisation of the racing line for each might end up slightly different if you recalculate them, but I would be surprised if it made any noticeable difference.

But if you want to do it, you can find the files in data\knw folder.

EDIT: I should explain, that the knw files mainly contain curves that were optimised to produce the fastest lap time for the given vehicle. They don't contain the actual speeds at each point on those curves. The speeds are worked out each time you enter the game and also as you drive around, while the car changes a bit. So that's why it doesn't make much difference to replace the knw files for small changes in AI prediction. Maybe if you have a favourite mod/track combination, it could be worth deleting the relevant knw file but I guess for most people it's not worth redoing all of them.

Bingo!

Seems fully sorted in the combos I have tried. Good stuff!!

It looks like AIs in D43 are more willing to keep their distance from each other instead of being very interested in overtaking others in D42.

Not sure if it's the new braking calculation, but AIs do braking better than before.

I've always tested the combo of TBO+LX+FZ5+RAC+FF GTRs with road_super in WE1 since D40. Faster cars start from behind.

I found a plausible issue with AS4 track for AI... there is chance that AI seems to not choosing to drive full speed on straight after corkscrew section. I did found out now a several cars now. If this is related to tyre physics, then it is fine.

EDIT: It happens on slower cars almost everytime, while faster cars it is dependable for other reasons of why it is not occurring exactly everytime.

EDIT: I am on D42, forgot to actually DL and install D43, mb.

Can you give a specific example of a mod that drives too slowly there?

Quote from Scawen :

Can you give a specific example of a mod that drives too slowly there?

Coming up pretty soon! with .spr file. Producing --->



EDIT: On this .spr file, it happens right before the corkscrew, and on the end of .spr. Seems like it is always a little bit different each time, but on here, this Hadraplan reproduces this issue 100%.

Another example, although much rarer, is LX8 GTR Evo made by Blz0. There it is reproduceable, but i have only managed to see it three times... it lasts from 1 second to 1 fraction of the second, everytime different.

Then there are ofc other vehicles, but have seen only just once.


EDIT: I try to catch more evidence.

EDIT4: Seems like this may be involved as well as what kind of setup AI is currently using, I am using ofc Level 5 AI, but just a little change to something, such as gear ratio, and then it is not happening if at all. This is absolutely interesting, so I will test this on rather bigger scale. Also using full grind instead of just 1 vehicle.