I think that AI's in lfs use deep neural network algorithm to learn how to drive, but do not quote me on that. In that case *.knw files store the biases and weights for the neuron layers. What I'm trying to say is that they are a able to learn, so you can race against them and teach them how to pass. But as you say, the problem is in the starting point and that is a default path. The good news are that in new patch lfs can generate AI paths automatically. In that process you can see how the AI path generation is itterative and in each step the AI's get better and better track times. Further improvement is possible only if human players is racing against them and they get pretty decent actually, at least for the beginner racer.
So AI's in lfs being a low priorrity is a huge understatement. There has been a lot of progress and developement in that respect already.
Yeah, it looks like you spent a lot of tris on the engine model, oh that is unfortunate
I think it is worth simplifying it, you did a great job with the mod in general and I'm sure it took a lot of time to create it in such details, but compromises must be made.
You are obviously clueless of how much work is required to "add rain". It's not gonna happen any time soon, probably never, devs have much more important thighs to finish first.
Not counting the regulars, most player in demo servers are most of the time barely keeping the car on track even with perfect weather conditions, with rain it would be a complete crashfest.
You have to showcase the menus and how simple the interface is. Another thing you have to show the demo multiplayer scene and tight evening racing at [AA] Blackwood GTi server, or other MRc servers. Ridiculous fps rates and also the mods, the continued development and some progress report pictures with new lighting, mention that new physics is on the way. I also like how chat is done, that is a big part of community, you can chat even while racing with wheel for example.
While we are here, I would like to also point out another very important fact about LFS. Again one of the old remnants, due to the capabilities of very early game control hardware.
All axes are scaled into integer numbers from -1000 to 1000, giving a tiny 2000 steps of resolution in any axis. This is completely fine for pedals, where most commercial wheels have about 256 (Logitech) or 1024 (Thrustmaster) resolution steps. Some fanatec or similair load cell pedals can have 4096 total steps. But for steering axis most wheels have full 16 bit range and that's where the problem may appear.
Full axis resolution supported by direct input API is 16bit, or total 65536 steps. When you reduce this to only 2000 steps, this is already coming to a point where you are deprived of small adjustments in steering input. Again this may be fine in the current FFB physics engine of LFS, where even if you had more resolution for steering input it would not matter, because the engine is probably not able to calculate forces down to this much precision anyway. Note that a direct input API supports only -10000 to 10000 for the magnitude of FFB effects (it is not even the full 16bit range). In this regard, the devs have already significantly improved LFS to support full FFB magnitude and 100Hz refresh rate. Before the FFB magnitude was +-200 and refresh was 25Hz, which was ok for early 8bit FFB wheels.
I consider this as a major point that needs to be fixed when the new physics is released. The game has to be able to utilize all the features of the FFB wheels, at least in the sense that it is able to utilize and resolve all the available steps in all reported axis of a wheel. Details of how the FFB is calculated and the required precision is a huge topic, that I do not want to go into at this point. I leave this up to devs to decide. So far, the overall character and feel of FFB in LFS is in my opinion by far the best, compared with any other more modern sim. I would definitely like that new physics engine tries to keep that feel and only add small details resulting from increased calculation precision, or calculation rate, or numerically simulating some parameters that were previously hardcoded or loaded from some approximated tyre slip tables.
I will repeat myself - adding a way to visually monitor FFB forces inside the game in realtime is a must with new physics engine. For example, even as simple graph such as this is a great feature to have.
You are absolutely right. Well, I believe this is some remnent of the very early development of LFS (started in 2002), when FFB has just been around very shortly (introduced 1999). This setting has not been changed since. I guess devs gave themselves a bit of the headroom. At the time there were barely any FFB wheels and all had very low res 8bit FFB, which still does not explain why they scaled it that way, but it probably had to do with the inability of FFB wheel to reproduce any low forces, so there was a compromise to clip everything from mid to high forces, such that you can have some feel in low forces.
FFB strength in LFS before clipping for some cars gets as low as 35%. The game offers up to 200% FFB strength and this is ridiculously high as pretty much all forces are clipped. So values from (35-200)% are absolutely not usable for any wheel and mind you this is >83% of the wasted adjustment range. Even with T300 or G29 I'm using (8-21)% depending on a car and track combo.
There are two solutions:
[1] make FFB strength parameter into a floating-point number (if it is not already) and allow adjustment of at least up to 2 decimal places, or just simply rescale the parameter by multiplying with about 5.0 internally.
[2] after reworking the physics engine, please do some rescaling of FFB strength such that for most of the cars on average the clipping point is at about 100% FFB strength.
well, as it is now, the devs did nothing to solve this problem yet, but the plan is to start working on new physiscs, so ffb interface might be revisited and you may be able to use that wheel in lfs
+1 for this, I would love to see +-1 increments instead of +-5. I always need to enter the options screen when I want to do small ffb strength changes during a race and that is not convenient, especially because the ffb is lost and the menu text makes me crash due to poor visibility 7 times out of 10
Tnx Scawen for the updates. Please pul the plug on the mods-related updates, it is a rabbit hole, it's time for physics (another major rabbit hole).
If everything fails, here is what I did. For a particular track just take the closest standard car to the mod you want to AI and copy its kwn file and rename it to the mod name which you can find as a skin ID. Then AI's will have a path but they will have to re-optimize it, so expect some crashes at some turns, but after a few laps, they should be ok. If you race against them that would be even better, because they can learn some new stuff when you are overtaking them.
Unplug the wheel and try different usb port. Do not use any third party usb hubs, connect straight to USB 2.0 motherboard port. Unplug any other game device before plugging back in the G29.
It happened sometimes to me that in G-hub the wheel was not recognized, even though startup calibration procedure went well and there was a windows sound of new device plugged in.
It's getting much better man, well done. The dashboard, steering wheel and gear stick are great. I would like the roll cage to be a bit higher, just above the windscreen, otherwise, it will not be safe in case of a crash, the shattered glass may hurt the driver. It will also help by not obstructing the view from inside. If you can angle the front roll cage bars a bit to be parallel with the side windscreen frame, that would also help.
If you can add a configuration without a roll cage would be a nice touch. Some more detailed interior textures would be nice as well. You can move the entire dashboard even more up such that everything is visible with default view and fov.
About the engine, would be good to tune cam timing a bit such that it has more torque at higher revs and that it can be a better match for FXS or XRT. Do not worry too much about matching it with our LX3 in multiclass server, it needs to be a competitor in TBO class.
Actually, LFS has a very good gamepad support. You can do everything that is physically possible, except for speed sensitive steering. Either use nonlinear steering with a full lock or reduce lock and use linear steering which I recommend.
The best way to use gamepads in LFS is just to reduce the car's wheel lock. For example XFG has 30deg max lock and if you set it to about 12deg or so and do not use wheel turn compensation, you will have a precise and linear steering. You have to experiment for each car and track, pit exit may be more difficuilt and if you lose control, there will be little to no chance to do big corrections, but that is the price you have to pay if you use 8bit (logitech) or 10bit (xbox) controller axis for steering.
If you open up that 3D printed box, you may find out that it contains Arduino Micro + ADS1115 module. Or it may have some other MCU like esp32 or stm32.
Does it show up in windows game controllers? In device manager, does it show up as a virtual COM port?