• Welcome to PiBoSo Official Forum. Please login or sign up.
 
July 03, 2020, 10:18:12 AM

News:

GP Bikes beta18c available! :)


Message For Piboso And Devolved Converstion

Started by Stout Johnson, June 15, 2020, 08:22:45 PM

Previous topic - Next topic

Stout Johnson

June 15, 2020, 08:22:45 PM Last Edit: June 15, 2020, 08:38:24 PM by Stout Johnson
Dear Piboso,

I haven't had a look at a version of GPB for quite some time, but out of curiosity I installed beta18b. There's a lot of things I like but also a few things that are really frustrating.

Pros:
+ nice new addition of Assen (could use a bit more detail though, especially grandstands)
+ nice Moto2 bike (engine sound could be better though)
+ sightly improved tire simulation
+ nice rider animations and hangoff styles

Cons:
- even on moderate downshifts (nowhere near high revs on downshifts), the rear locks up; the rear even locks up under braking when there is NO downshifting at all. this is not realistic at all - is this a broken slipper clutch simulation?
- the tires heat up very strangely and become almost unrideable after 2 laps; only way around that is to brake and downshift ultra conservatively and try to keep revs on downshifts under 40% of max rev which is absolutely unrealistic
watch this video for example on how aggressively real downshifts are made without rear locking up

- tire temps and grip simulation is too crude and artificial: tire grip out of the pits is best, after a few laps on track tires tend to slide very much; so in GPB with cold tires the bike has perfect grip and with warmer tires the bike has less grip and tends to slide out rather randomly - very strange; in reality it is the other way aroud
--> I know this is supposed to simulate overheating of tires, but it is broken; first there is too much grip with cold tires and the way the tires overheat is exaggerated in GPB; in reality overheating tires do generate less grip than optimal tire temperatures, but still more grip than cold tires. overheating tires do generate a bit less grip and the tires do spin a tad bit more (so accelerating is a bit slower) and the tires slide a bit more under ultra-heavy braking (so one has to brake a bit earlier), but they just do not magically slide or spin out like in GPB; in GPB when tires overheat a bit, the bike even tends to slide out with almost no lean angle; also in GPB you can brake latest when shifting down ultra conservatively (ultra late) whereas in reality you can brake latest when shifting down aggressively and using the engine brake power of high revs.
- in GPB the grip simulation has a tendency to be only on the extremes: either there is absolute perfect grip or no grip at all and the bike sliding like crazy

Overall I still like GPB a lot: I like the suspension simulation, the overall feeling of the brakes in GPB. That is what still creates the "itch" I felt when I discovered GPB more than 10years ago when I did lap after lap after lap trying to master the bike. But at the moment that joy goes away rather quickly because the tire simulation seems to be rather crude in terms of tire temps and grip simulation. Also the way the rear locks up under even moderate braking and downshifting is absolutely a turn-off and makes riding long stints with realistic braking impossible. At the moment I "enjoy" getting out of pits and using the grip of the cold (!) tires for one lap, after that it becomes a sliding festival or trying to "walk on eggs" with ultra conservative braking and downshifting and trying to have the tires in "non-icy icy sliding mode"... which is annoying and I quit after 15mins.

@Piboso: If you find the time, I really would like to hear how you explain the differences in what your sim does and what happens in reality on braking and downshifting (as shown in the video I posted above). Is there possibly a problem with the slipper clutch simulation? And where and how are you trying to improve GPB in terms of tire simulation?

Kind regards
Stout
    -----------   WarStout Kawasaki Team   -----------

Vini

I agree, the M2 has a few problems.
That said, mod bikes have surpassed the default bikes handling a long time ago.

Stout Johnson

June 16, 2020, 06:12:13 AM #2 Last Edit: June 16, 2020, 06:15:00 AM by Stout Johnson
Probably Vini, but it is sad that a bike "simulator" hast such massive flaws out of the box. Even I - as a long year GPB enthusiast - am deterred that the bike created by dev of the sim does not reflect realistic behaviour. I can only assume what people might think that try the sim for the first time. It has to be hard for newbies, because of the physics and it for sure takes a learning curve for every simulator. But the underlying motorcycle physics need to be on point, especially for those who seek a simulator. Those who are interested in GPB over stuff like Milestone Games are for the most part people that have much real-life riding experience, to a fair share people that have raced in real life and therefore know how a bike and tires should behave. And I am pretty sure GPB with the M2 has a lot of potential to attract, but as it is now will rather do the opposite. Also I think for modders, it would be good to have reference for tire sim and how it is meant to be.

I would be much interested in where the big boss stands on this.

Quote from: Vini on June 15, 2020, 11:03:04 PMI agree, the M2 has a few problems.
That said, mod bikes have surpassed the default bikes handling a long time ago.
Which mod bikes are available for b18, which one can you recommend?
    -----------   WarStout Kawasaki Team   -----------

PiBoSo


The M2 engine braking curve needs tuning for sure.
Also, maybe the slipper clutch simulation is not fully correct yet.

As for the stock tyres...
They are a long-standing problem that will have to be addressed sooner rather than later.
There are a couple problems:
1) There is too much grip when the tyre is slipping at the max lean angle
2) However, when the grip is lost, it drops too quickly, making it almost impossible to control the bike
The algorithms are correct, it should be a simple matter of tuning the tyres' physics data.
I think that some mods already did a good job improving the tyres' behaviour.
"Obviously your ambition outweighs your talent".

h106frp

I think the slipper is working quite well, the only thing that is not obvious is the basic definition of the spring strength, is it related to the max_torque?
i.e. a low power/torque bike would use light springs and less plates but a big powerful bike would need stronger springs and more frction area (plates). So it is not so obvious how to control the rate of pressure plate displacement.

The generic behaviour of the tyres can be tuned but I would be interested in comments on my post as how to define the temperature/grip/wear behaviour.

Stout Johnson

June 16, 2020, 12:14:13 PM #5 Last Edit: June 16, 2020, 12:23:45 PM by Stout Johnson
Quote from: PiBoSo on June 16, 2020, 10:33:11 AMThe M2 engine braking curve needs tuning for sure.
Also, maybe the slipper clutch simulation is not fully correct yet.

As for the stock tyres...
They are a long-standing problem that will have to be addressed sooner rather than later.
There are a couple problems:
1) There is too much grip when the tyre is slipping at the max lean angle
2) However, when the grip is lost, it drops too quickly, making it almost impossible to control the bike
The algorithms are correct, it should be a simple matter of tuning the tyres' physics data.
I think that some mods already did a good job improving the tyres' behaviour.
First off, thank you for responding Piboso. Also thank you for your honest response. I salute that, very professional on your part.

Looking at what you write and also seeing what h106 (who has a ton of knowledge in that respect, I conversed with him privately on that matter) wrote, I think in all likelihood the slipper clutch simulation might not be main problem, but rather the engine braking curve. That would also explain why it happens even with no downshifts at all, only braking. Even then the rear locks up rather easily.

The other big problem is the simulation of the tires. First off, the relation between temps and grip needs to be realistic.

Secondly, the simulation of the tire temps needs to be realistic. As far as I know you account for 3 different areas of the tire areas ("temperature rings") which provide different heating areas - center, left, right? This is much too crude from my point of view.
Even for cars, tire simulation could be more sophisticated, because even there the stress on the tires changes rapidly. But for cars, the tires at least are on contact with the asphalt almost all the time. For bikes, depending on the lean, some areas of the tire are very far away from the contact patch and the surface and are rapidly cooling when not "under stress".
You can see how fast tire temps on the surface vary, increasing almost immediately under stress and cooling almost instantly when stress decreases. Even for cars with tyres acting almost everytime as a contact patch this is very obvious. This should be even more obvious with bike tires as there is only a small contact patch and therefore only a narrow contact patch ring during a certain lean angle. The majority of a motorcycle tire (most temperature rings) is not acting as a contact patch and therfore for the most part cooling. For bikes, the cooling should be much more rapid compared to car tires - for example when leaning from the left and then to the right because then the left flank would be far away from that contact patch, away from the asphalt and getting almost no stress (except in the deep areas of the tire due to macro-deformation) and would cool off in the surrounding air.

If you would implement a very thorough tire temperature simulation with numerous temperature rings, you might get rid of many problems with grip simulation. You might even be able to simulate high-siders - which is very crude at the moment and always comes very much delayed at the moment in GPB. It would be the ultimate testament to your simulation if you could get that simulated.

And having many different contact patch/tire temperature rings might help simulated it much better. The area under extreme stress while spinning (overheating and spinning) gets less stress if the rider releases throttle and the spinning disappears, the tires cools off very fast on the upper surface and the rubber does not melt anymore but becomes firmer again. All of a sudden there is an immediate increase in grip. Also the lean of the bike may change during that process and the contact patch is now on a cooler "ring" (a different area of the tire where there had been less stress and therefore less heat) - that significant rise in grip creates inertia in the tires which works its way up over the bike into the rider and propels him off the bike. As stated above, for all that to be realistic you also need a realistic connection between tire temps and grip level as asked by H106 here --> relation between temps and grip.

Talking about all of this, one thing comes to my mind. Some years ago, when you implemented the simulation of rev changes with lean due to the circumference differences of a tire (increasing wheel velocity with higher lean due to smaller circumference), you implemented some sort of circumference measuring in GPB. Could that possibly also be usable for the problem of differing "contact patch/tire temperature rings"?

I think it would not be too hard to implement? Of course you'd have to simulate tire temps for each "ring". But maybe there do not need to be too many rings. I'd say having like 11 rings (left, ride) per side and one wider ring for the center of the tire might be sufficient? If the center ring is being used for 5° of lean to each side (so 10° l/r overall). That would leave ~ 55° of lean left for 11 rings. That would account for one temperature ring per 5° of lean. That might just be sophisticated enough for relatively realistic tire temps and therefor calculation of actual grip, depending on the lean and the tire temps of that tire area that is being used for a certain amount of lean. That might even make the sim less dependent on trying to tune the current tire physics data because the differing tire temps and therfore grip levels per lean would make it more realistic in that respect already. After all, it might be impossible to tweak tire physics data and expect realistic behaviour (e.g. realistic slides, high-siders) if the tire temps model is so crude as it is now. What do you think?
    -----------   WarStout Kawasaki Team   -----------

Myst1cPrun3

June 16, 2020, 02:28:17 PM #6 Last Edit: June 16, 2020, 03:00:03 PM by Myst1cPrun3
The tyre heating behaviour for most of the riding process is set in the tyre files by the mod creator.

In there the tyres can be set to heat and cool differently, as well as separate values for being on the sides of the tyre for long periods of time.

To my knowledge no-one has got tyres right, as quite frankly, there is very little information available. Sure we've got pacejkas formula but to me that's all there is.

As h said with the post that was  referenced there is very little, perhaps no information on how this affects tyre behaviour.

I cannot state however about how the tyre is split or how to implement a more accurate model as I don't know. I don't actually know how inaccurate the current one is as like I said no one has got a tyre right.

(Some feel good to ride on, but when out on a different bike, they feel awful. Now this could be a bike issue I don't know, however to my knowledge the bikes are accurate using real data so there is no reason for it to be the bikes if that makes sense. Whether it's an issue somewhere else again I have no clue)

As for the m2, I also share the issues with it, as do most users of gp bikes it seems.

It's why I have decided to attempt to 'redo' various parts of it, in an M2 V2 mod available on my Discord server. (Will be on forum's once complete) it's not finished yet, as the tyres need a huge update, (about 50% completed) so it's still a wip.

I also have a HP4 Race which is again on my discord server if that's your cup of tea. It is a first mod for me so while I tried to keep values as close to real as possible there are some concenssions, and it isn't as good as it could be I don't think.

While I personally aren't keen on them, There are new 500cc GP and MotoGP mods available, again on discord, and while I again am not a fan (seem unrealistic to me) I would suggest giving them a go as people seem to like them, and there does seem to be some care taken for their development.

One recurring theme here is Discord.

Even if you hate the program I reccomend joining the GPB Discord, as that's where most of the 'community' is these days, and due to restrictions on mods here on the forum's, most of them get released there first.

Hope I confused the matter even more  ;D





h106frp

Nearly all tyre files are copy and paste of the default, the heat rate and wear are adjusted to suit the mod bike (mainly down to chassis weight and the fact that the default tyres are off by quite a bit) Fiddling the couple of values to make the rear slide works but its a bit off from reality .

What we need is to fix is the instantaneous dynamics of the tyre grip - its part of what makes bike riding a challenge and PiBosos sim different to the scripted behaviour found in most bike games.

Hawk

Quote from: Myst1cPrun3 on June 16, 2020, 02:28:17 PMEven if you hate the program I recommend joining the GPB Discord, as that's where most of the 'community' is these days, and due to restrictions on mods here on the forum's, most of them get released there first.

Hope I confused the matter even more  ;D

It's NOT "MODS" that have been restricted on the forum, It's links to "ILLEGAL" MODS that have been restricted; two totally different contexts mate. ;)

Anyone is free to post & link about their MODS if they have not been ripped, copied, or used & altered without any permissions from the original creator.

So feel free to post your "Legal" MODS here in the forum. It's not a problem at all. ;)

Vini

June 16, 2020, 07:08:11 PM #9 Last Edit: June 16, 2020, 07:24:47 PM by Vini
@Stout: Manu's MotoGP bikes are by far the best handling bikes out there at the moment.

Some mods have a really realistic front with proper chattering and "progressive grip loss" under heavy loads etc. The M2 is lacking very much in that regard in addition to the engine brake / slipper clutch problems.
The rear tyre of most mods on the other hand could still improve further. Especially powerslides are not smooth/progressive enough. Loss and regain of traction occurs too abruptly. I don't think this is purely due to tyres, though. The virtual rider may simply be too tense on the steering, not allowing the bike to work freely.

Quote from: Myst1cPrun3 on June 16, 2020, 02:28:17 PM(Some feel good to ride on, but when out on a different bike, they feel awful. Now this could be a bike issue I don't know, however to my knowledge the bikes are accurate using real data so there is no reason for it to be the bikes if that makes sense. Whether it's an issue somewhere else again I have no clue)
It is the bikes. No bike is accurate, as there are a couple of free parameters like chassis flex or steering forces for which no empirical data exist.

Myst1cPrun3

Quote from: Hawk on June 16, 2020, 06:43:39 PM
Quote from: Myst1cPrun3 on June 16, 2020, 02:28:17 PMEven if you hate the program I recommend joining the GPB Discord, as that's where most of the 'community' is these days, and due to restrictions on mods here on the forum's, most of them get released there first.

Hope I confused the matter even more  ;D

It's NOT "MODS" that have been restricted on the forum, It's links to "ILLEGAL" MODS that have been restricted; two totally different contexts mate. ;)

Anyone is free to post & link about their MODS if they have not been ripped, copied, or used & altered without any permissions from the original creator.

So feel free to post your "Legal" MODS here in the forum. It's not a problem at all. ;)

kinda is the same thing though, as to my understanding this covered all things not scratch built, whether the converter has permission or not was not relevant to my knowledge. But I could be wrong. IDK

Either way, the fzr750 is the latest mod on the forums, and before that it was 2018 the last mod, the rest are on Discord, so its still well worth having.

Quote from: Vini on June 16, 2020, 07:08:11 PM@Stout: Manu's MotoGP bikes are by far the best handling bikes out there at the moment.

I would have to disagree with that. For 2 main reasons.

1) I highly doubt anyone here has ridden a 2019 MotoGP bike, so no-one can really say what is realistic.
2) 0% to 100% throttle at max lean doesn't kill you, but is the fastest way around the turn... No bike, Grand Prix Prototype with all the fancy electronics or not should be capable of this, and the onboard data from the real bikes certainly goes against this also.
But again, this has been covered extensively on the discord, so I won't go further. (and its not what this thread is about)




Quote from: Vini on June 16, 2020, 07:08:11 PMSome mods have a really realistic front with proper chattering and "progressive grip loss" under heavy loads etc. The M2 is lacking very much in that regard in addition to the engine brake / slipper clutch problems.
The rear tyre of most mods on the other hand could still improve further. Especially powerslides are not smooth/progressive enough. Loss and regain of traction occurs too abruptly. I don't think this is purely due to tyres, though. The virtual rider may simply be too tense on the steering, not allowing the bike to work freely.


This is 1/2 true to my knowledge. There are some good tyres out there, (F+R) but again, no-one has got them 100% right. The best way I've found is to have a value where the front tyre can have a higher maximum angle than the rear, as the width difference seems to result in the front 'going off the edge more' at max lean. Allowing the Front that bit extra angle over the rear makes the whole experience, particularly on corners with camber changes (MG Hairpin) a whole lot more confidence inspiring, without harming the realism, as the bike will only lean as far on the joystick/input as the tyres minimum - max lean if that makes sense. (The tyre with the lowest maximum lean dictates the total max lean of the bike, but if the camber changes, that max lean gets exceeded. Its why lean is limited on grass, but if you are beyond that lean limit entering grass you fall)

The M2 is lacking in that regard also, and the EB and slipper clutch aren't right as mentioned. (Which is a WIP on the M2 v2 I'm doing)

The rear tyres overall are tricky, and as you say, no-one has that right at all I don't feel. Some are closer to others, but it really is a guessing game it seems.

As for the VR, the VR is editable in the bike CFG. How far this goes, I have no clue, I haven't experimented enough, so there may be an underlying issue with it. But I did reduce a value to 0 and he lost all input on the bars, so it is possible to have him relax more in the bike CFG's so its a mod thing mainly to me.

Quote from: Vini on June 16, 2020, 07:08:11 PMIt is the bikes. No bike is accurate, as there are a couple of free parameters like chassis flex or steering forces for which no empirical data exist.

Yes and No. While there isn't any concrete data as to actual values, there is data and formulae on the internet to help workout values that are roughly in the ballpark.

This I think needs Piboso to clear up on just how he gets his values, and what formulae are used to calculate it

Myst1cPrun3

PS:

Note on the tyres:

I did find through my testing that the front being 'dragged' around the turns can be reduced or removed by reducing the 'slidiness' of the rear tyre. Bit odd that it had this effect I know, but I will do some more testing for the new tyres on the M2V2 to consolidate this theory.

Vini

June 16, 2020, 08:21:45 PM #12 Last Edit: June 16, 2020, 08:23:59 PM by Vini
Quote from: Myst1cPrun3 on June 16, 2020, 07:48:57 PM1) I highly doubt anyone here has ridden a 2019 MotoGP bike, so no-one can really say what is realistic.
2) 0% to 100% throttle at max lean doesn't kill you, but is the fastest way around the turn... No bike, Grand Prix Prototype with all the fancy electronics or not should be capable of this, and the onboard data from the real bikes certainly goes against this also.
1) How the bike "feels" under your ass is irrelevant anyway since you are not playing the game through a motion-simulator that accurately simulates all forces acting on you.
In general, the mod looks and sounds very close to real MotoGP footage when pushed to the limit.
2) That is a horrible criterion to judge the realism of a mod as the ECU simulation is very bad in GPB in general. Also, real MotoGP riders can go to full throttle a lot quicker than you might think and if they wanted to, they could easily set up the traction to allow for instant flat-out but of course this would result in slow times.
Riding aids (including low direct lean values) in general mask a lot of the true bike handling in GPB. Going insta-flatout on the MGP bikes will definitely "kill" you without TC.
In any case, while the mod is obviously not perfect, it's the most realistic of the mods we currently have. It even managed to improve the downhill understeer to some degree.

Speaking of that...
@PiBoSo: What is the consensus on the downhill problem? What is the cause and could it be fixed in the future? This is the one big physics engine problem remaining in the game.

Quote from: Myst1cPrun3 on June 16, 2020, 07:48:57 PMAs for the VR, the VR is editable in the bike CFG. How far this goes, I have no clue, I haven't experimented enough, so there may be an underlying issue with it. But I did reduce a value to 0 and he lost all input on the bars, so it is possible to have him relax more in the bike CFG's so its a mod thing mainly to me.
Which value are you talking about specifically? I guess the problem is the lack of dynamics in the steering forces: While you obviously need a lot of steering input to change direction at high speed, they should actually be close to absolute 0 under heavy acceleration (corner exit). My analysis and proposed experiment regarding this has still not been adressed by PiBoSo.

Quote from: Vini on June 16, 2020, 07:08:11 PMYes and No. While there isn't any concrete data as to actual values, there is data and formulae on the internet to help workout values that are roughly in the ballpark.
I don't think those results are anything close to accurate. There is a reason simulators are not a thing in MotoGP like they are in F1. Modern chassis design is pretty much black magic, as the Ducati and Yamaha mishaps of the past have shown. Lastly, there is no "virtual rider algorithm" in reality.

Myst1cPrun3

Quote from: Vini on June 16, 2020, 08:21:45 PM1) How the bike "feels" under your ass is irrelevant anyway since you are not playing the game through a motion-simulator that accurately simulates all forces acting on you.

Agreed

Quote from: Vini on June 16, 2020, 08:21:45 PMIn general, the mod looks and sounds very close to real MotoGP footage when pushed to the limit.


Looking at the inputs and how they differ I have no choice but to disagree to this. The inputs are not representative of the real bikes when compared side by side. The visual behaviour of the bikes is also not representative when compared to real life.

Quote from: Vini on June 16, 2020, 08:21:45 PM2) That is a horrible criterion to judge the realism of a mod as the ECU simulation is very bad in GPB in general. Also, real MotoGP riders can go to full throttle a lot quicker than you might think and if they wanted to, they could easily set up the traction to allow for instant flat-out but of course this would result in slow times.


The ECU simulation is bad in GP Bikes, however the 'harshness' of its intervention is user controlled.
As you said they could do such a thing IRL, and it would make the bikes slower, but, in the game its the fastest way to do so.


Quote from: Vini on June 16, 2020, 08:21:45 PMIn any case, while the mod is obviously not perfect, it's the most realistic of the mods we currently have.


Gonna have to disagree again. I mean  just remembering back to the Valencia session we had the other day, the fastest setup was maximum damping on one end, minimum on the other, and making the bike as long as possible to increase its steering speed. This is COMPLETELY Backwards, and highlights massive issues with the mod values.


Quote from: Vini on June 16, 2020, 08:21:45 PMSpeaking of that...
@PiBoSo: What is the consensus on the downhill problem? What is the cause and could it be fixed in the future? This is the one big physics engine problem remaining in the game.

+1
Quote from: Vini on June 16, 2020, 08:21:45 PMWhich value are you talking about specifically? I guess the problem is the lack of dynamics in the steering forces: While you obviously need a lot of steering input to change direction at high speed, they should actually be close to absolute 0 under heavy acceleration (corner exit). My analysis and proposed experiment regarding this has still not been adressed by PiBoSo.


spg0 = -340
spg1 = -2
sdg0 = -100
sdg1 = 0.5
sig0 = 0
sig1 = 0

These are the values that seem to edit the VR values to me. There is a thread on these but the reply from Pib was 'It is too complicated to explain' which I can only assume is why he's not bothered to reply...

Quote from: Vini on June 16, 2020, 07:08:11 PMI don't think those results are anything close to accurate. There is a reason simulators are not a thing in MotoGP like they are in F1. Modern chassis design is pretty much black magic, as the Ducati and Yamaha mishaps of the past have shown. Lastly, there is no "virtual rider algorithm" in reality.



If Piboso has used formulas to calculate the physics of GP Bikes, (which he has) then there is no reason it should have changed for this area. Its why I want Piboso to explain how he calculated the values, as this will allow us to better calculate ours.

It is a black magic art IRL, but using material properties and formulae using real world data, we can get a very good guess as to various values. We have real values to go on or get close to for the most part so there is no real reason for it to be 'black magic' here

h106frp

@MysticPrun
Are you going to try adding ECU controls to you M2V2? I never really tried the M2 but after Stout raised the question I had a look in the .cfg and after observing a lot of forum discussion I was surprised to find it has no ECU modes other than pit-limiter. Everything on the M2 is basic mechanical at the moment.

I can understand why PiBoso has the bike this way for development as it makes it much easier to get the mechanical side correct id you do not have all the electronic fudgery going on masking basic physics issues.

A quote I found on the web for M2 ECU standard;
the initial electronic features that will be available for the Moto2 riders were confirmed as torque maps, engine braking and launch control.

But as rumoured, the most well-known electronic rider aid, traction control, will not be present.

"Traction control will not be introduced for next season," said MotoGP Director of Technology Corrado Cecchinelli. "It doesn't mean it won't be in the future.

So you could define the engine maps (torque), braking and launch - the old mura990 is an example of the script.