I have not had time yet to make a minimal example, but I've noticed a case where it commonly happens. If I create a robot that has only axial joints it seems to happen more frequently. https://imgur.com/a/eUqgudR https://imgur.com/a/eUqgudR This would make sense as the solver has no degrees of freed...
Sorry, I don't at the moment. The code is quite complex. I'll see if I can save the robot configuration that causes the bug and write a minimal working example.
I have some deadlines at the moment so it might be a little while before I can provide a minimal example.
This is my workaround code: function checkNil(nanArray) if nanArray then for i = 1, #nanArray, 1 do if nanArray[i] == nil or nanArray[i] ~= nanArray[i] then return nil end end return nanArray end return nil end I just wrap every call with this function. I assume it would be faster to do this in c++ ...
This is a pretty minor bug as I can work around it, but just thought you may want to know. If I call sim.getConfigForTipPose(), most of the time if no solution is found it returns nil. However, sometimes it returns: {-nan,-nan,-nan,-nan} I noticed this because I was accidentally setting the joint po...
Thanks for the patch, I've been running with new version of the legacy remote api plugin. It has failed to connect as before, but I want to perform more tests before I say for certain. If I set useAlternateSocketRoutines to true, does that have exactly the same behaviour as the unupdated plugin? Tha...
It does not crash overnight but as I mentioned before this alternative connection method is unreliable and causes other problems for running parallel instances.
I have not run it for long enough yet to see if it solves the main problem, but it causes other problems after 1 iteration of my algorithm because the connection time because highly unstable. Spawn 7 instances of vrep launch the optimiser that connects to those 7 instances and computes a function in...
I don't run out of memory, but after 1600 seconds I cannot connect to the remote API client. I last ran this on v3.6.1 rev2 so things might be different now.