Development Media

So... sometimes it's easy to overlook issues in the technical design even if you thought about it for a long time...

As it turns out, I completely forgot to consider slowmotion when designing the H-Pattern shifter movement a while ago.
If you manage to physically shift faster with your input hardware than the shifter can move around in jbeam, it will then take weird paths and get temporarily desynced from the internal state.
This is somewhat tricky to do when playing in realtime, but trivial to encouter when the game is in slowmotion.

To fix this without turning the solution into a closed-loop system (which has its own set of issues and difficulties), the only somewhat easy to implement solution is queueing.
So with the next update, I've changed the system to record all the shift movements and execute them in order. If abused, this leads to funny video content like this:

Video - Click to Play - Direct Link

 

Chillin

 

After a bit of time creating something, time for one of the best parts... Testing how it works!

Video - Click to Play - Direct Link

 

A little update on the shifting stuff
Not everything will be finally implemented once it releases, more of a first iteration, waiting for feedback from you and manual tuning for various different vehicles.

 

Fired up! _
#TeaserTuesday

 

I'll just leave this little clip of a development test structure here without an explanation

Video - Click to Play - Direct Link

Hint: Its main use won't be in a vehicle with an engine/motor.

 

Random dev scribbles or "why simple school math can be relevant for your job"...

 

Teaser Tuesday is back!

(see full size image here)

 

Slide that speculation into the Update Speculation Thread

Video - Click to Play - Direct Link

 

Emerging
#TeaserTuesday



(full size w/o branding here)

 

Smashing debut!
#TeaserTuesday

 

Rollin', rollin', rollin'
#TeaserTuesday

 

Sliding right into your feed!

It's #TeaserTuesday

 

Some insight into one of the "invisible" features of our simulator, force feedback for all you steering wheel drivers!

While working on further improvements to FFB code, there's wide variety of factors that we need to care about.

For example, how does the new candidate FFB code behave at 30fps vs 60fps vs 400fps. How does it respond to suspension geometries from half a century ago vs modern suspension designs vs a pigeon front axle which was engineered to be asymmetric because why not. What about gear/belt/direct-driven steering wheels. Or casual street driving vs spirited driving vs driving at the real limit. Drifting vs gripping. Doing donuts. Parking a truck. Jumps. Stunts on 2 wheels. A smooth racetrack surface vs touges vs dirt trails or sand. A car equipped with a quick vs slow steering rack. Powersteering installed or not. Different ingame strength settings, different ingame smoother settings. Different hardware steering rims. Different manufacturer software settings.

As you can imagine, the list of factors goes on and on. In an ideal world, you would want to test all factors combined in all possible ways, to verify things are still working fine no matter what any random person will do in BeamNG.drive. This is called a "combinatorial explosion", and in practice it's impossible to test in its entirety.


To attack the problem, we try to automate testing as much as possible. However, FFB is one of the few simulator subsystems that requires interactions with the real world (with a real USB steering wheel).
This means that only a tiny part of the tests can be automated, while the rest relies on real human beams at the steering wheel.

The automated tests can be ran in a matter of minutes, which is nice. But the manual tests can take hours/days very easily.

This is the result of a single automated test. This one measures high frequency vibrations suffered by the steering rack:


And here's one of the manual tests (those we do by actually driving the sim):


[if you ever noticed and wondered about the "Max Steering" value shown in the "FFB Graph" UI app, shown in the screenshot above, then now you will know: I added it quantify those typical oscillations you've probably noticed at some point, and which are natural in a delayed-feedback-loop such as FFB in computer sims]

In many other cases, the result cannot be quantified with numbers on the screen. Instead, we are forced to resort to a subjective judgement of the driving feel. In some case, the improvements can be so small that you have to double check back and forth, see if they are within placebo effect, or if it was a real improvement. This can be a difficult journey but is always worth the effort, because the sum of such tiny incremental improvements is noticeable when checking against the previous publicly-available version of FFB.

While testing various candidates of FFB code, specially during subjective testing, you want to be able to switch between them all very quickly. This keeps the impressions fresh in your recent memory, and makes comparisons easier. To help with this, I have this simple selector program, where you can pick any FFB version. This is what it looks like:


Despite all the complexity and difficulties, we do our best and we never stop trying to improve FFB. We hope to be shipping further improvements this year, if our ongoing FFB research leads to positive results. So far it's looking promising, and we have some FFB candidates that greatly improve over what's publicly available in v0.32 - but you never know if you'll find some odd corner case that takes you back to the drawing board, so we won't promise anything. Our fingers are crossed tho!

If you've read this far, and on a more entertaining final note, here's a bonus video: one of the manual tests I sometimes do, it helps gauge the responsiveness of FFB at low speed. And I guess it also helps to check the collision model of that lamp post

 

Shiftergates of hell

 

Aaaand working

Video - Click to Play - Direct Link

 

#TeaserTuesday

 

#TeaserThursday

Disclaimer: We're adding a few new body parts! (not a remaster)

 

Testing conventional components with unconventional approaches

#TeaserTuesday

 

No gear too low, no hill too high!

#TeaserTuesday