B7 Mura 990 Small lean angles create large lateral oscillations

[h106frp]

I have tested a lot in first person and continually observe that when the bike is at very low lean angles under acceleration lateral oscillations set in. Increasing the lean slightly, raising the bike to vertical will cause them to cease.

Very noticeable of very long curves where you have to hold slight lean angles but you can also sense it on the rumble every time the bike passes through the low lean angle to a higher angle.

Not sure if its related but there does seem an issue with calculating the rear suspension mechanical damping;
setting 'RearLatDamping' to a lowish value (2 or 3) gives rise to huge non resolving oscillations even with the bike just rolling slowly in a straight line with no power.
setting it 20 seems to give a reasonable damping and a much reduced tendency for oscillations to build at low lean angles, going to 30 (over the top - i know) the bike explodes on the grid as soon as you leave the garage and 'core.exe'.

The 'FrontLatDamping' seems OK and you can set a low value with no problems

[CapeDoctor]

i've experienced this as well - possibly why i spend much more time on the Varese than the Mura....

[h106frp]

I spent the evening tuning the Yohji V1 config and its improved the situation greatly - its a fun ride now especially in 1P, much less prone to the shakes. You do need to try a few setups with the Yohji model as the garage settings are more effective for each different track but give it a go - not many bikes to try with B7 anyway

[HornetMaX]

What do you call "lateral oscillations" exactly ?

Because RearLatDamping does not relate to the rear suspension, it's the chassis/swingarm flexibility model damping.

[h106frp]

I know - which is why it seems so strange. The lateral oscillations can start at a very small lean angle and just keep building. You can take the same turn, different line with silly throttle settings and it will be OK though if you are carrying more lean.

Try setting RearLatDamping to 2 or 3 if you want a laugh - barely moving and the bike shakes itself to bits and does not decay - does not seem possible

[HornetMaX]

It is very possible: if one changes (significantly) the bikes structural parameters, the virtual rider may no longer grant stability.

[Napalm Nick]

I dunno what causes it but I can concur it is very annoying. Lots of people online talking about it. I think we all wait for B8 with fingers crossed.

[h106frp]

It is very possible: if one changes (significantly) the bikes structural parameters, the virtual rider may no longer grant stability.

Very evident on the stock bike since b7 and its horrible - it is what set me on my current quest 

Try the first straight on spa after eau rouge with stock 990. Riding  the long gentle curve in the track, tiny constant lean leads to big vibes. 
and exiting onto the last long back straight bit near what looks like a lake - same problem

It just feels like a bug somewhere in the system.

[BOBR6 84]

Bit like the.. exit the pit garage, turn left into pit lane, Wobble wobble wobble.. Straighten up away you go.. Thing.

[BOBR6 84]

Almost like the virtual rider is confused.. Steer, no.. counter steer. Steer, counter steer.

At very low speed the bike seems to lean instead of just turning. Could it be related somehow??

[HornetMaX]

At very low speed the bike dynamics are very different and it is very likely that the virtual rider is not doing the right thing.
I don't think it's a big problem as it is unrelated to the issues we experience at higher speed.

It could probably be solved at the price of some additional complexity in the virtual rider (like a dedicated low-speed controller plus some smart-ass way to switch between the low-speed and high-speed controller).

[Blackheart]

I have not read everything, but you refer to it? Just happened in the Monza Race tonight...

https://www.youtube.com/v/3J9N0am_dY8

[h106frp]

At very low speed the bike dynamics are very different and it is very likely that the virtual rider is not doing the right thing.
I don't think it's a big problem as it is unrelated to the issues we experience at higher speed.

It could probably be solved at the price of some additional complexity in the virtual rider (like a dedicated low-speed controller plus some smart-ass way to switch between the low-speed and high-speed controller).

I think something might have been tried already, been playing with the config and increasing kdamping1 hugely improves low speed virtual rider control but also seems to confuse him (overreacts) at speed and under braking (which is odd). spg0 effects low speed sensitivity, more negative = more sensitivity, spg1 effects high speed sensitivity the same way. I have jiggled it so its reasonably balanced but its not 'right' under braking as the rider seems to be the input to the oscillation. Still trying to figure sig0 and sig1.

@ Blackheart - the shakes look familiar 

[HornetMaX]

At very low speed the bike dynamics are very different and it is very likely that the virtual rider is not doing the right thing.
I don't think it's a big problem as it is unrelated to the issues we experience at higher speed.

It could probably be solved at the price of some additional complexity in the virtual rider (like a dedicated low-speed controller plus some smart-ass way to switch between the low-speed and high-speed controller).

I think something might have been tried already

No, I don't think it has been.

been playing with the config and increasing kdamping1 hugely improves low speed virtual rider control but also seems to confuse him (overreacts) at speed and under braking (which is odd). spg0 effects low speed sensitivity, more negative = more sensitivity, spg1 effects high speed sensitivity the same way. I have jiggled it so its reasonably balanced but its not 'right' under braking as the rider seems to be the input to the oscillation. Still trying to figure sig0 and sig1.

spg0, sig0 and sdg0 are the parameters of the PID controller stabilizing the "target lean to steering" part of the bike.  The "1" gains (spg1, sig1, sdg1) are the gains that adapt to speed, e.g. spg = spg0 + spg1 * speed (same for sig and sdg).

Kyaw, Kdamping0 and Kdamping1 are (probably, has we have no official confirmation) the gains of the "yaw controller", the thing that tries to have the bike "tracking correctly", i.e. keep neutral steering behavior, no oversteer, no understeer.

From what I've seen, you'd better focus on the spg, sig, sdg gains and leave Kyaw, Kdamping0 and Kdamping1 alone (even if you put them at zero, the bike is still rideable).

But knowing that is not enough if you don't know what you're actually trying to control.

Tuning this by trial and error is borderline impossible: the two controllers interact, the bike characteristics vary with the bike speed, the bike has a "non minimal phase" behavior (I won't go into details but essentially the fact that to lean left you have to initially steer right, it makes things harder), the non-minimal phase thing is there only above a certain speed (that's why at very low speeds the current virtual rider sucks). It's a complex control problem: even taking a huge simplification (linearize at different bike speeds and try to find an LPV controller that stabilizes all the resulting systems) is hard without proper tools.

I don't know how Piboso does all this, but if I had to do it first thing I'd do is to buy a Matlab license (with a few toolboxes on top).

[Vini]

I don't know how Piboso does all this

Sadly, probably not as you would.



Anyway, I also think that the virtual rider is the source for most physics problems we have currently (apart from the missing swing arm pivot adjustment on the stock bikes).