(Not sure if this was posted but…) I have heard by several users on ROBLOX, that Legacy mode will be removed. That will hurt every single roller coaster/ ride based game on ROBLOX. The roller coasters run on friction, on the old physics engine that the new engine can not replicate. There is no known way to fix this as of yet, causing all coasters to suddenly stop moving on the track when it hits as something as simple as a turn, or a steep hill. It would be sad to see a place like my own ROBLOX Point: Theme Park, to longer work after 2 years of hard work to finish it.
We are aware of those side effects of the new physics engine. The issue is that it attempts to keep everything as rigid as possible contrary to the old physics engine that was based on springs.
However there are possible solutions. Very soon we will be releasing new constraints that include springs and velocity motors. These will allow you to build suspension (for the roller coaster cars) that allow some flexibility.
I’m pretty sure he was talking about the legacy combination, that is, Spring solver and Legacy properties. The physical properties update has made some major changes to how parts and especially hinges behave in spring physics, and setting density to 1 isn’t a fix.
Which things have you tried that do not work? Have you tried reverting densities to 1? Does it work? Are we talking about specifically Physical Material Properties here, or is this somewhat PGS related. I have multiple scripts in the threads announcing this that show how to revert to old behavior when using the new system.
@chefdeletat I think people are talking about hinges with respect to super-large mass ratios.
@NWSpacek Setting Density to 1 is actually a fine fix in some cases, although it shouldn’t be needed.
@khanovich it isn’t really a fine fix. You can see that in legacy properties the locomotive is relatively rigid. In new properties with density of 1, the locomotive is really loose and flops all over the place.
Please note: I am not advocating for the removal of Legacy. Do not misconstrue this solution as being pro-physical properties.
I took a look at this roller coaster because it is easy to manipulate the properties of, and it has been confirmed to be reliable across physics engines.
Using Legacy physics as a benchmark, I tried 7 different combinations of Friction, FrictionWeight, and boost velocity. The data was collected using a 1/10th second loop, recording the velocity, and plotting each data point on a graph. The vertical lines in the data are only indicative of rounding pixels in the GUI and not of any errors in the experiment. The velocity is displayed as a vertical bar. Each horizontal pixel represents 1/10th of a second, and each vertical pixel represents a single stud per second. Please note that the data is plotted based on the actual time and not the simulation time.
It’s kinda garbage data with all of the different tests cluttered together, so I will show you the Legacy run and the run that mimics Legacy the closest. The colors have been changed to increase contrast.
The following changes were made to get the behavior as close as possible to Legacy:
Friction of wheels set to 0 (0.3 in Legacy)
FrictionWeight of wheels set to 20
Friction of boost sections set to 1 (same as Legacy)
FrictionWeight of rails set to 20
Velocity of boost sections decreased to match Legacy boost speeds (in Legacy the train did not reach the full velocity of the boost section).
Elasticity of wheels and rails stayed the same at 0.5.
EDIT: there is no need to change the FrictionWeight of anything, you just need to set the wheel’s Friction to zero. Rails that already have friction will still be able to propel the train. For some reason, the train loses just a fraction of a percent less speed than Legacy, which compounds over time: