AI Improvement Plugin by JR

Plugins AI Improvement Plugin by JR 1.2

Login or Register an account to download this content
If you're willing to experiment a bit and try to test to see if AI fuel strategy is working, that would be great...

If you do some experimentation and notice anything interesting, please do report back here! In fact, regardless, I'd be interested in hearing about your AI racing experience with the plugin.
I think the plugin is managing AI fuel strategy and driving behavior. Using the analyser, I can see that the AI are refueling <100% but enough to get them to the end of the race which is good.

I turned off most of the ballast penalties so cannot comment on those aspects of the plugin. I run with the blue flag ballast penalty but I must admit, I have not noticed this making a difference.

I can't get the other aspects of the plugin to work, such as the live championship viewer.

I do however, have another issue with the AI pitting needlessly without taking fuel or tyres. I've started a new thread for this which has some image outputs from the log analyser: AI pitting for no need
 
Last edited:
Thanks for the testing feedback! Really appreciate it. :)

I think the plugin is managing AI fuel strategy and driving behavior. Using the analyser, I can see that the AI are refueling <100% but enough to get them to the end of the race which is good.
This is good. Only thing is that I think the AI does already intelligently choose how much fuel to put in at the start of a race, even without the plugin – I believe it's based on the FuelUse value in the track's AIW file. So I don't think it confirms, in and of itself, that the fuel strategy piece of the plugin is working as intended.

I turned off most of the ballast penalties so cannot comment on those aspects of the plugin. I run with the blue flag ballast penalty but I must admit, I have not noticed this making a difference.
To test if blue flag behaviour is improved, I'd do this. Choose a multiclass mod, select a track with long straights, run a session, and ride with a high class AI car watching them navigate through traffic. In particular, watch out for if they get "stuck" in a straight line behind slow cars. Do this with and without the plugin, and compare how they behave.

You could also do a single class race (e.g. F1) and simulate races going long enough that the leaders catch lapped traffic. Similarly, ride along and watch the leading AI and slow AI, do this with and without the plugin, and try to spot any differences.

When doing this, do be mindful that it can be easy to have a placebo effect and spot something that is not there. Heck, I've sure done it. So be mindful and observant. ;)

I can't get the other aspects of the plugin to work, such as the live championship viewer.
Me neither. I found that most of the visual widgets are broken nowadays. This worries me – why are they broken? And does this mean other non-UI aspects of the plugin are also broken?

I do however, have another issue with the AI pitting needlessly without taking fuel or tyres. I've started a new thread for this which has some image outputs from the log analyser: AI pitting for no need
Yeah, good idea to put that into a new thread. :)
 
Last edited:
To test if blue flag behaviour is improved, I'd do this. Choose a multiclass mod, select a track with long straights, run a session, and ride with a high class AI car watching them navigate through traffic. In particular, watch out for if they get "stuck" in a straight line behind slow cars. Do this with and without the plugin, and compare how they behave.
I have run some multi-class endurance and 100% race length F1 races, and I am not convinced the blue flag behavior is changing with and without the rF AI plugin.

Therefore, I am not convinced the plugin still works as originally intended.
 
I have run some multi-class endurance and 100% race length F1 races, and I am not convinced the blue flag behavior is changing with and without the rF AI plugin.

Therefore, I am not convinced the plugin still works as originally intended.
Thanks so much for the testing and feedback. :) While (as you can see in my earlier posts above) I was initially quite bullish on the straight-line multiclass overtaking and blue flag behaviour via the plugin, I now fear I was simply experiencing a placebo effect. All too easy to do. :(
 
Last edited:
Hi - I have continued to run with/without the plugin to see if I notice a difference.
I think there is a noticeable difference with the AI and how they race the player. They tend to give a bit more room in my opinion. Without the plugin, the AI seem much more likely to pretend you're not there and turn in on you. I'm not sure any of the functions in the Race.ini file work or make a difference to the plugin functionality.
However, I have noticed an issue with the plugin. When running with some of the newer mods (F1 RMT 2011, WEC by Simtek and WEC 2015-16), the frames drop drastically and the game is unplayable. The only way to recover in-game performance is to remove the AI improvement plugin. Having said that, the frames are fine when using other "heavy" mods such as CTDP 2005/2005, Enduracers and CARTfactor with the plugin. I'm not really sure why the plugin works with some mods and not others, what's causing the frame drops, or how to fix it...
 
Last edited:
Hi - I have continued to run with/without the plugin to see if I notice a difference.
Thank you so much for doing this! :thumbsup:

I think there is a noticeable difference with the AI and how they race the player. They tend to give a bit more room in my opinion. Without the plugin, the AI seem much more likely to pretend you're not there and turn in on you.
This is great to know. Suggests there are some improvements to overtaking logic that are still functional.

I'm not sure any of the functions in the Race.ini file work or make a difference to the plugin functionality.
It's been a while since I tested the plugin (been busy in ACC, AMS1. and RBR) but I believe that's what my testing led me to believe as well.

However, I have noticed an issue with the plugin. When running with some of the newer mods (F1 RMT 2011, WEC by Simtek and WEC 2015-16), the frames drop drastically and the game is unplayable. The only way to recover in-game performance is to remove the AI improvement plugin. Having said that, the frames are fine when using other "heavy" mods such as CTDP 2005/2005, Enduracers and CARTfactor with the plugin. I'm not really sure why the plugin works with some mods and not others, what's causing the frame drops, or how to fix it...
My gut reaction - do you know if the problematic mods feature 3D models ripped from other titles? For example, I recall GTR 2 WEC mods using models ripped from the mobile game Real Racing or from Assetto Corsa. And even if they aren't ripped, are they very high in detail?

Why I ask is because one of the plugin's features is to improve collision detection. I'm wondering if these mods have poorly optimized (or entirely unoptimized) 3D meshes used for collisions in rF1's physics. In other words, I'm wondering if the plugin is trying to calculate "better" collisions, but it is using a ton of computational power to try to deal with these overly detailed 3D models.

For context - collisions in rF1 (and other sims) are not necessarily determined by the object's model itself, but a simpler "collider mesh" associated to each object. This is why, for example, tire barriers usually seem flat when you crash into them - because their collider mesh IS a flat box. Cars in well-crafted mods (like CART Factor or CTDP F1) will all have specially-designed simpler meshes used for collision physics calculations... hence why they are not causing you any trouble. But low effort or ripped mods may not.

IF this is what's going on, it would confirm that the plugin is improving collision detection! Which would mean this annoying bug is, in a way, a good thing.
 
My gut reaction - do you know if the problematic mods feature 3D models ripped from other titles? For example, I recall GTR 2 WEC mods using models ripped from the mobile game Real Racing or from Assetto Corsa. And even if they aren't ripped, are they very high in detail?
I cannot say for certain and I'm not sure how I could check the collision meshes to confirm this.
However, your theory seems very credible and I would not be surprised if you are correct
 
Last edited:

Latest News

Online or Offline racing?

  • 100% online racing

    Votes: 93 7.7%
  • 75% online 25% offline

    Votes: 129 10.6%
  • 50% online 50% offline

    Votes: 173 14.3%
  • 25% online 75% offline

    Votes: 341 28.1%
  • 100% offline racing

    Votes: 473 39.0%
  • Something else, explain in comment

    Votes: 5 0.4%
Back
Top