Thanks for taking a look at aerodynamics! We tuned that aircraft model so it is easy to fly at lower speeds, like below about 400 km/h. You can try altering the center of mass and position of the elevators to improve responsiveness, but the aircraft will become less stable the more responsive you make it.
Having said that, I think what you are seeing in the video is related to the high speed the aircraft is at (~1400 km/h, above Mach 1). It looks like you increased the max thrust? Our aerodynamic model is a compromise between real-time performance, robustness (i.e. doesn’t crash), and believable behavior. What your example shows is some behavior that can occur when the forces are so large they may cause the underlying numerical method to be unstable. For flight vehicles it really works best at lower Mach numbers. However, thanks for this feedback! We can use your example to find ways to expand flight regimes the model reliably, and in real time, works with.
Thank you for the detailed explanation! It would be fantastic if a universal solution could be found for this. I’m working on a flight simulator where I plan to include not only very fast aircraft, like the one from Top Gun: Maverick that reached Mach 10, but also futuristic models with extreme maneuverability and speed. As it stands, I’m concerned that these concepts may push the current aerodynamic model beyond its limits. I’m excited about the potential improvements, and I’d be happy to continue providing feedback to help refine the aerodynamic model for more diverse flight regimes.
Thanks for trying out aerodynamics! Your feedback will help us improve the model.
Torque roll is a realistic effect and its impact is related to the amount of torque needed to drive the propeller (so its moment of inertia and drag) and the moment of inertia of the aircraft. To counter act the roll you’ll need to add a torque in the opposite direction using control surfaces like ailerons or a counterrotating propeller (see the KSP video). You could, for example, use a FluidForceSensor on the propeller to measure the torque it’s applying and then use that information to automatically trim the aircraft via ailerons.
Do you have an example we can look at? So far we have not seen an issue with propellers so long as the masses of the propellers and aircraft are scaled appropriately. However, if do have an example that is particularly bad we’d love to take a look so we can improve the model.