Better Failure Analysis

[almagnus1]

One feature that would help us figure out (and fix) RUD situations is the ability to be able to replay, and watch, a failure from an arbitrary point of view, and be able to slowly step through the failure as it occurs.

Ideally, being able to see aerodynamic and physical stresses on the parts would help, as that can help us determine if we need to add more struts, SRBs, fins, duct tape, and/or prayer to fix the design flaw.

I mean, I don't mind a spectacular RUD every now and again, but being able to pull more data from the fireworks show would help improve that aspect of the gameplay experience.

[Evanitis]

Well, I like field testing until I find the problem...

But... replaying the explosions in slow-motion from various camera angles? Good idea. Yes please!

[smjjames]

I totally agree here. While the F3 menu does help show where the failure started, it doesn't show or tell how it happened.

[nikokespprfan]

Yes agree, this is a game where one learns from ones mistakes, this would be a helpful tool for that.

[KrazyKrl]

This sounds like one of those features, which are amazing, but are horribly complex to code.

I'm all for a replay system in KSP, implementing it would be a huge problem though. You would need to save every physics tick to get an accurate replay of what happened, because of floating point inaccuracies in the physics engine. You would need to save these during one of the most intensive periods of a crafts flight (collisions and destruction); this is assuming saving physics ticks for replay is even possible in a Unity game. Saving rendered frames would be easier, but with massive disk space cost; and probable performance loss. Saving just graphics brings a few other problems, inability to interact with the scene for one. I guess you could annotate the saved video with event log stuff.

 

Yes, it would be amazing; and I would love this feature. The work behind a replay though... It might be easier to use something like NVidia Shadowplay to save videos yourself as a workaround.

[almagnus1]

20 hours ago, KrazyKrl said:

This sounds like one of those features, which are amazing, but are horribly complex to code.

I'm all for a replay system in KSP, implementing it would be a huge problem though. You would need to save every physics tick to get an accurate replay of what happened, because of floating point inaccuracies in the physics engine. You would need to save these during one of the most intensive periods of a crafts flight (collisions and destruction); this is assuming saving physics ticks for replay is even possible in a Unity game. Saving rendered frames would be easier, but with massive disk space cost; and probable performance loss. Saving just graphics brings a few other problems, inability to interact with the scene for one. I guess you could annotate the saved video with event log stuff.

 

Yes, it would be amazing; and I would love this feature. The work behind a replay though... It might be easier to use something like NVidia Shadowplay to save videos yourself as a workaround.

If we know position and velocity vectors, wouldn't that be enough to replay the failure?

For most failure scenarios, it's because there's a weakpoint in the design (like oscilation caused by having too small of a node bear too much stress), and most of those failure scenarios wouldn't need the precision that would make floating point rounding an issue.