Continue simulation based on saved states

[smhadisadati]

Is there a way to stop the simulation and save the states and close the program, then open the scene again and load the latest states as initial conditions (position and velocities) and continue the simulation?

[fredroy]

Hi @smhadisadati

It may be possible to do what you want to achieve with the ReadState/WriteState.
These 2 components are supposed to read/write states (x, v, f) associated to each timestep of your simulation.
You may do the trick by editing the output file from ReadState, changing the last time step to 0 and removing all the previous lines. But this is not really the design of these 2 components so I am 100% sure it will work 🤨.
And of course the time-step accross your stop/start simulations will change.

[hugtalbot]

As we also discussed together @smhadisadati, several internal mechanical information remain not accessible (as Data) in SOFA. Therefore, this prevents from saving / exporting these info (internal stress), i.e. it prevents from restarting from a pre-stressed configuration.

A possible but brutal approach was used by @p-shg for reinforcement learning method by using .... os.fork but it works.
This topic is strongly related to the SOFA Data, which could be an very interesting technical work in the core of SOFA to finance.