Physics cleanup

I am not sure how realistic the game physics is supposed to be. But there are many many things that are slightly wrong and for submitting fixes I need to understand the roadmap.

1. Time scale in the movement code. x+=100*dt*dx is almost a literal quote. Then people wonder why you need to divide theoretical orbital speed by 10 — that's why, that 100 effectively changes the timescale and lowers G.

2. Various cleanup like the fuel_to_stop formula (which happens to be wrong). This should be fixed anyway, but it is frustrating to fix it without fixing the basic movement logic.

3. Acceleration cap. It seems to be intended for flying inside the star disk, but there are planets with orbits affected by the acceleration cap. And it caps absolute value of acceleration instead of degree of non-uniformity. Not to start with the basic idea of controlling for energy…

4. Unequal time steps and basic precision optimisations… I think that enforcing a fixed time step in physics (even if wall clock time between frames is slightly different), calculating all the forces before moving anything and alternating the order of «change x» and «change dx» can do quite a lot for numerical stability without costing much in performance.

And in the ideal world it would be great to be able to accelerate time.

But to properly fix physics there has to be some decision about what parameters are supposed to be the tunables.

I hope that the magic constants like 100 can be eradicated — but this requires rebalancing the default time scale.

I also hope you won't object to changing the acceleration cap…

Whether you want precision management cleanup — I don't know.

Thanks; 1 and 2 seem like pretty straightforward fixes.

For 3, if you can find a better way to avoid the crazy "shoot off into the outer reaches" bugs without an accel cap, I'd be willing to hear it. Fixing the time step alone won't do it though. I vaguely recall implementing fixed time step a while back and it not helping anything, but maybe my implementation was flawed. If you want to give it a shot, I'd be up for it.

Edit: as for tunable physics "constants", the only one I have in mind is the time factor; you can see it calculated and applied in the update function in main.lua.

You probably consider G an acceptable target for tuning if gameplay requires it. Of course, G and timescaleand space scale have very similar effects.

Fixed time step is nor for acceleration cap replacement, but it sometimes improves other things.

OK, I take that as an agreement to rebalance the scales without changing the core physics. I mean, I am not going to do a good job at rebalancing the plot timescale, but the physics changes will require something to be done.

Ouch. Currently the universe has such small distances that the time actually has to be slower than the real time.

And the relative weights of planets seem to be a crime against science on their own.

By the way, if you want auto pilot to be required, there is an option of making stars actually heavy, planets somewhat heavy, and stations almost-realistically light. Then the precision required to get into a stable coordinated orbit with a station would need either huge time slowdown or some good autopilot with real orbital mechanics calculations.

OK, my current situation is: removed the cap and the main victim is trajectory. If I go into a small orbit around the space station (impossible with the low acceleration cap), it obviously goes crazy. The normal flight is actually fine because I increased precision just a bit without significant performance problems, but trajectory is expensive…

Started in !23 (merged)

mentioned in issue #123 (closed)

One problem here is that players with saved beta-1 games will see very unusual behavior if they upgrade; planets and portals will suddenly fall into the sun. We need some kind of in-game migration mechanism to re-seed the position and velocities of things when it detects you're running a game from an earlier version.

I've added checks for loading beta-1 games that recalculate all the positions of the bodies in the current system.

mentioned in issue #143

Is there any reason to specify portal triggering time in terms of outside universe and not the in-game universe?

No, I think it's just that the portal activation code predates the time factor stuff.

I have a feeling that I have seen an asteroid with the center of mass on the surface. Don't know if a one-off glitch or a bug.

FWIW I've implemented fixed timestep on master now.