Jump to content Jump to content

AI ridiculously fast in high speed corners


Recommended Posts

The AI is not always driving all the time. That's for all sim racing games though, not only F1.

Sometimes the AI is just generally following a set of parameters. A script. That's why on some games you will bump the AI on a corner and it will not bulge or quickly retake its line like if it was nothing.

What @Blackbird is probably suspecting is that the set of parameters are off for some sections of some tracks. This take a good time to identify and fix as... Well, let's say you're cannot ask your actor to simply adlib a scene,  you're having to rewrite the whole script.

  • Like 1
Link to post
Share on other sites
10 hours ago, Sparticus_19 said:

Can you explain?  I legit am curious by this.

When coding AI speed in general at some stage of its development something went wrong, particularly the AI behavior on most high speed corners on the game was broken as a result.

Link to post
Share on other sites

AUSTRIA - Spielberg

Hi there, today my analysis in Austria and yes, the AI seems totally on another planet in T3, after the study on Monza I didn't think the situation was so serious, honestly 😕

Once I'm completely off the pace I use only my data of Time Trial (in any case better than my Q attempts) and mainly the current TT WR data (from Playstation), here the data:
- from my side I'm more or less able to match the AI 105 in S1 and S2 while I'm 0,5 s. slower in S3, which is 0,7 s. in comparison with AI 110
- the TT WR is almost able to match AI 110 in S1 and S2 but it's also 0,5 s. slower in S3, no story at all.

tab.jpg.d185712cf691acbf06133c09973344ad.jpg

 

The telemetry graphs speak for themselves:

in the (5) Wurth I'm already in trouble while the WR man is still able to match the AI 105 (but not the AI 110), impossible for anybody to keep up the AI speed at the (7) Rindt and in the following short straight to prepare the (8), unfortunately here something looks really broken with the AI 😢
 

bestTT_vs_AI110.thumb.jpg.2adcb9e6320240e11e4f3f4aa6b69800.jpg

bestTT_vs_AI105.thumb.jpg.762714ea0bb0ab51d95529230d4c70fc.jpg

WR_vs_AI110.thumb.jpg.8ed68948eeab767682c2b01cfe63c444.jpg

WR_vs_AI105.thumb.jpg.125ef5f8d139087b4a06afb2582db1ca.jpg

 

Bye,
Nuv

 

 

 

Edited by Nuvolarix
  • Like 1
  • Thanks 1
  • Sad 1
Link to post
Share on other sites
32 minutes ago, CityGhostGamer said:

Well, T7 feels essentially the same, I have to break very early to make the corner, otherwise I understeer. So no apparent changes to handling in that regard. To be fair, patch notes didn't say anything about high-speed corners, only low-speed. So there's that. Now let's see how AI handle it.

Nope, AI behaves the same way, 'red' after patch 1.08

Image1.thumb.jpg.d25412c5c7af6b8f9c4d1bae914577c7.jpg

Edited by Nuvolarix
  • Thanks 1
Link to post
Share on other sites
  • Administrator

Just to confirm here, this is NOT in Patch 1.08. User has been contacted spreading wrong information.

UPDATE: To clear users name here, was a honest mistake. Another website spreading false information. The best source for all legitimate patch notes is me. If it doesn't come through the Codemasters official channels, don't trust it. 👍

  • Thanks 3
Link to post
Share on other sites

Hi there, when I did my last telemetry check in Austria, where I was so off the pace, I used the WR telemetry instead and that gave me an idea to speed up everything: I do not need to train each track myself, I can just use the WR and compare it with the AI!

 The results are interesting and quite clear (and hopefully useful to developers).

In the following posts I will (this part of the message will be repeated at each):



  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

SPAIN - Barcelona

T3 & T9 [KO]

1858435296_04.Barcelona_WR_vs_AI110.thumb.jpg.655f7575893616cc2e3ea52ebb713b8e.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

GREAT BRITAIN - Silverstone

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T1 (Abbey) - T9 (Copse) - T14 (Chapel) - T15 (Stowe) [KO]

21394429_10.Silverstone_WR_vs_AI110.thumb.jpg.8ec75c57715fcd35b66ecb9883070b17.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

BELGIUM - Spa-Francorchamps

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T12 (Pouhon) [KO]

762973877_12.Spa-Francorchamps_WR_vs_AI110.thumb.jpg.de9339399bd479989f5ca8fbc66a2a94.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

BELGIUM - Spa-Francorchamps [WET]

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

Eau Rouge + Radillon [KO]

1781389861_12.Spa-FrancorchampsWET_WR_vs_AI110.thumb.jpg.b78bcd837ed7be95902453faa4fe5e26.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

The Netherlands - Zandvoort

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T7 - T8 [KO]

1896708628_13.Zandvoort_WR_vs_AI110.thumb.jpg.e4c69ba689074b97da54b1fc117c65bd.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

JAPAN - Suzuka

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T4-T5-T6 ('S' curves) - T8 (Degner) - T15 (130R) [KO]

1121094811_17.Suzuka_WR_vs_AI110.thumb.jpg.3e9b3fe1eb1773cf9d8039baf88738fc.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

USA - Austin

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T3 - T4 [KO]

2126647994_18.Austin_WR_vs_AI110.thumb.jpg.e20d4a20abc8d7ae77b98316f11f82f9.jpg

To better show the track positions which include (only) T3-T4:

track_position_1.thumb.jpg.a3c222852909ff7ceeecceda16a54c37.jpg

track_position_2.thumb.jpg.2a14a216a93ec5dda80b05fa2bf0dd56.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

As last I summarize integrating @BarryBL's table with the telemetry results obtained:

Image2.jpg.25b712c07c080f3e891640e3c79ba705.jpg

In the next days, the check on the other tracks will follow (at least on 'dry') to verify that there are no other AI problems somewhere.

Greetings,
Nuv

Edited by Nuvolarix
  • Like 1
  • Thanks 2
Link to post
Share on other sites

Hi again, I’ve worked a little bit more on the matter by completing the telemetry study on ‘dry’ at each track. Most are ok, but I have found also new interesting data which I’m going to publish in the next posts related to 5 further tracks/curves in order to help the devs to find a fix or at least ‘where’ a fix is necessary.

First is

HUNGARY - Hungaroring

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T11 [KO]

1371113450_11.Hungaroring_WR_vs_AI110.thumb.jpg.ec310ba0a71e3c8acd7d0f3d4642e7fd.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

MEXICO - Mexico City

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T9 [KO]

592094942_19.Mexico_WR_vs_AI110.thumb.jpg.67d5d6ebfd20e16b2e1a1a800ddd18f1.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

BRAZIL - Interlagos

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T6 (Ferradura) - T7 (Curva do Laranjinha) [KO]

1824264547_20.Interlagos_WR_vs_AI110.thumb.jpg.a5ccc6fd4b8a7b286953be68bac4aa54.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

AUSTRALIA - Melbourne

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T11-T12 [KO]

1319531966_21.Melbourne_WR_vs_AI110.thumb.jpg.0268617d2bc153eb022cd11b1ab6701c.jpg

Greetings,
Nuv

  • Thanks 1
Link to post
Share on other sites

CHINA - Shanghai

  • compare the Playstation WR record with 'no aids' VS a qualifying lap of AI 110 in all the tracks reported in Barry's table (Note: the WR is the current best possible, the AI Q lap is typically made in a single session or at most two, in some cases it can be improved and quite easily achieved, but still good for telemetry purposes);

  • use red circles in the points of attention;

  • sometimes the WR has an 'alternate' line (I don't know if for connection problems or other), when I realized it I took another similar time, but in essence the general line does not change and the comparison is still possible;

  • in some tracks there's a much higher top speed by AI, this is due to setup reasons for which I ignore it.

T7 [KO]

1196994869_24.China_WR_vs_AI110.thumb.jpg.6469d9dcfdda6f4f61a33e8dafcf79d9.jpg

Greetings,
Nuv

  • Like 1
  • Thanks 1
  • Agree 1
Link to post
Share on other sites

As usual I summarize using the @BarryBL's tab with the new data as well:

Image1.jpg.e275ad5304237ac2293581c74a7ed1b1.jpg

All the tracks have been checked with the telemetry on ‘dry’ while I’m not going to work on ‘wet’ (hopefully they are already ok) unless devs would need it or someone else would ask for a specific test.

Ciao! ☺️

  • Like 1
  • Thanks 2
  • Agree 1
Link to post
Share on other sites
On 8/10/2021 at 3:05 AM, blackheartbart7 said:

I put this graph in another post re AI in the wet - it probably has some meaning here as well.

 

Attached is an image of a race telemetry, Human in green AI in brown. Where we brake, turn and throttle (see green lines in the graph) - the AI does micro brakes and micro throttles every millisecond (see the up and down brown lines) and so they don't suffer the physics model we do when cornering. 

I don't know why they have set the AI handling this way - it would make sense for them to mirror human actions. We can't manipulate things this way - and it definitely has an impact on handling. 

Maybe it's always been like this - I don't know. Only found this telemetry program a month ago. 

IMG_0133.jpg

Hey mate, can you tell me what is this telemetry program?

Link to post
Share on other sites
On 9/6/2021 at 9:41 AM, Nuvolarix said:

As usual I summarize using the @BarryBL's tab with the new data as well:

Image1.jpg.e275ad5304237ac2293581c74a7ed1b1.jpg

All the tracks have been checked with the telemetry on ‘dry’ while I’m not going to work on ‘wet’ (hopefully they are already ok) unless devs would need it or someone else would ask for a specific test.

Ciao! ☺️

 

Thank you so much for your great analysis @Nuvolarix!

I hope @BarryBL and Codemasters will take the data into account!

  • Thanks 1
Link to post
Share on other sites

@Nuvolarix WOW O_O

I'm speechless, mate. Top job, I applaud your devotion and determination. You rock! 🤘

I agree, Hungary T11, Mexico esses section and Australia high-speed chicane - all definitely are problematic as well. And I had no idea that AI brake so late in USA, that's mental 🤪

Barry has to take a look at this, this is some brilliant work right here. Codemasters should pay you for this 😁

Edited by CityGhostGamer
Link to post
Share on other sites
30 minutes ago, CityGhostGamer said:

@Nuvolarix WOW O_O

I'm speechless, mate. Top job, I applaud your devotion and determination. You rock! 🤘

I agree, Hungary T11, Mexico esses section and Australia high-speed chicane - all definitely are problematic as well. And I had no idea that AI brake so late in USA, that's mental 🤪

Barry has to take a look at this, this is some brilliant work right here. Codemasters should pay you for this 😁

Thank you for the nice words @CityGhostGamer! 😊

I'm just happy to help make the game better and better.  In my humble opinion it is really very well done overall and this anomalous behavior of the AI in some corners is just a pity, so hopefully Codemasters will be able to focus on the resolution rather than wasting time identifying the anomalous curves, which can also be done by an user.

Fingers crossed 🤞 😉

Regards,

Nuv

  • Like 1
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...