I'm not very sure if it's really related to 930b bit I didn't see this before. Please have a look in both videos below. Almost impossible to race against AI anymore, they're bouncing, jumping and turning over in a non-sense way.
Any track with high curbs (ISI tracks included) so it's apparently not a track issue. Now I'm going to test other cars series. There's something wrong with Meganes and Clios AI, they're doing so many mistakes even with AIMistakes=0 in PLR file. Examples: In IndianĂ¡polis Race Course, Turn 4 where is mandatory AI to brake in the middle of the corner, they spin almost every time, producing full-track yellow flags in sequence and make impossible to race. In Sebring, Meganes hit the ripple strips, starts to bounce in a "elastic bed" way and eventually flips. This issue seems to start happening with 930b. In Loch Drummond, in the fast chicane (photo), Meganes or turn over, or spin or go out of the track. In addition, there's a second strange effect: when the front AI runner spins, all those coming behind spin too, in a domino effect. Probably the cars "see" the incident in front of them and apply too much brake, spinning too. That always causes full track YF. I tried to give some setups to AIs in order to correct that, including extremely soft suspensions, but no effect, I'm wondering if the "AI SETUP" button is even working.
have you tried a default player.JSON? just to eradicate anything you may of edited and forgot about (happens to me all the time!)
I tried a fresh PLR but no luck. The good news it's look like to be a Megane's suspension issue. Please notice the instability of the car in last Sebring corner (already with new PLR): Now see the UDR cars at same place, much smoother (even bumping a bit): In rF1 I use to set AIMinPassesPerTick=7 in order to improve the AI physics at cost of some GPU demand. I don't know if this variable still exists in rF2 HDC files. On behalf of Clio's mistakes everywhere (despite the drivers names could suggest bad driving ), to me is a typical AIAimSpeedsPerWP variable issue. The Megane's behavior reminds me a no-spring car supported directly by the bumpers.
This is happening in the very bouncy bits of Sebring, may be something got uncovered and greatly exacerbated in this track. By getting it replicated with stock content it should really help.
Yes, Sebring is bumping but other cars can run fine. I've just found the HDV AI's of Megane has low values for coded AI suspension: AIMinPassesPerTick=6 // minimum passes per tick (can use more accurate spring/damper/torque values, but takes more CPU) AIRotationThreshold=0.20 // rotation threshold (rads/sec) to temporarily increment passes per tick AIEvenSuspension=0.0 // averages out spring and damper rates to improve stability (0.0 - 1.0) AISpringRate=0.4 // spring rate adjustment for AI physics AIDamperSlow=0.15 // contribution of average slow damper into simple AI damper AIDamperFast=0.50 // contribution of average fast damper into simple AI damper AIDownforceZArm=0.13 // hard-coded center-of-pressure offset from vehicle CG AIDownforceBias=0.0 // bias between setup and hard-coded value (0.0-1.0) AITorqueStab=(1.00, 1.00, 1.00) // torque adjustment to keep AI stable AIFuelMult=0.99000 // PLR file override for AI fuel usage - only positive value will override, see PLR for default AIPerfUsage=(1.00, 1.00, 1.00) // PLR file overrides for (brake power usage, brake grip usage, corner grip usage) used by AI to estimate performance - only positive values will override, see PLR for defaults AITableParams=(28000, 19) // PLR file overrides for (max load, min radius) used when computing performance estimate tables - only positive values will override, see PLR for defaults For example, 370Z (more stable): AIMinPassesPerTick=5 // minimum passes per tick (can use more accurate spring/damper/torque values, but takes more CPU) AIRotationThreshold=0.20 // rotation threshold (rads/sec) to temporarily increment passes per tick AIEvenSuspension=0.0 // averages out spring and damper rates to improve stability (0.0 - 1.0) AISpringRate=0.88 // spring rate adjustment for AI physics AIDamperSlow=0.35 // contribution of average slow damper into simple AI damper AIDamperFast=0.65 // contribution of average fast damper into simple AI damper AIDownforceZArm=0.10 // hard-coded center-of-pressure offset from vehicle CG AIDownforceBias=0.0 // bias between setup and hard-coded value (0.0-1.0) AITorqueStab=(1.0, 1.0, 1.0) // torque adjustment to keep AI stable AIFuelMult=-1.0 // PLR file override for AI fuel usage - only positive value will override, see PLR for default AIPerfUsage=(-1.0, -1.0, -1.0) // PLR file overrides for (brake power usage, brake grip usage, corner grip usage) used by AI to estimate performance - only positive values will override, see PLR for defaults AITableParams=(-1.0, -1.0) // PLR file overrides for (max load, It can be the source of strange AI behavior. But I'm noob on rF2 package / unpackage, it looks like I gonna read the book before to play around those variables.
I hope there will be a lot of extra liveries in that one!!! PS, now I'm on the livery subject, are you planning on doing something like you did with the howstons again??? I thought that would have been a nice addition for for instance the Clio's as well (those just need a very full grid IMO ) EDIT :: Oops, it seems I 'Bounced' off-topic a bit
Ah, apparently this issue is already gone with the Megane we have internally... so we'll try to get that out quicker.
Thanks Tim, please notice that I'm not saying there's something totally wrong with the car, they're just findings. Actually human vs human they are great, the probelm is when human vs AI. I'm a great enthusiast of Megane and Clio packages, they're are very fun. Maybe my next real life car be a Megane as result of rF2 driving lol
https://www.youtube.com/watch?v=lZSmUgJPOaY Edit the json file: AI listening Snoop Dog during races: false;
Hi, just want to report same problems with Megane & Clio. Spin over at LRP and Mid Ohio (e.g.) //Edit: oversee the Clio`s update. Downloading now... //Edit2: rechecked my Race Rig Installation. Clio v1.51 is still installed there. The spin over problems exist with v1.51, too
I'll try to run an AI Calibration in certain tracks today that I'm sure exactly where they'll spin and will report if it make some difference.
I confirm that AI Calibration improves the AI lap times in qualifying and testday but doesn't solve the AI lines issue during the race. For some reason the AI cars insist in to put one wheel out of the tarmac, apply the brake with one wheel on the grass and spin lap after lap, producing YFs in sequence. It seems to be something with defensive line or something damaging AI drive line. Cars runs perfectly when they're far one to another but spins every time in turn 4 when they're grouped.
Simple question: why in Earth AIs do different lines from the qualy / practise to race even when they're not being pushed?