Okay, so in my amateur code hackings with the Ardupilot codebase I have been trying to do some rate-based operations that need the loop time in the calc. I’m currently running SITL on a Ubuntu VM and X-plane on the base Win7 OS but putting some trace calcs in suggest that the loop time can be quite variable - enough to deviate significantly from the plane.G_Dt loop value which I assume most of the rest of the code is using.
Firstly, is there a simple way to improve the sim performance?
Secondly, could I push SITL off to a Pi or similar and connect to X-plane with a real-world ethernet connection so that I could have the code running on a separate uC? Presumably with a (more) dedicated processor running Ardupilot it may run more consistently?
I’m completely Pi-illiterate so this is a steep gradient for me. Just wondering if it may be worth the effort.
Thoughts appreciated!
The issue you’re running in to is probably because your front end and backend aren’t lockstepped, so can get out of sync - particularly if you’re graphics heavy. Try using a sim that supports lockstep scheduling. For Plane I can’t remember which option is best (I generally just use the basic inbuilt sitl), but from memory JSBSim is lockstepped.
Hmm. OK, thanks @james_pattison You seem to be suggesting that Ardupilot running in SITL mode is getting at least some of the timebase info from X-plane rather than just pseudo-sensor data. Am I on the right track here? I will have a dig about in the code…