Replies to This Discussion

Permalink Reply by hopslink on October 29, 2010 at 3:15am

AP does not derive dynamic response from a static model of your airframe (nor will any other generic autopilot), so it doesn't ask you for static parameters. Reason? - there is no simple generic static>dynamic model offering good accuracy for all airframes, so...

AP uses a general model for flight dynamics and you tune this to match your airframe's dynamic response by adjusting PID gains, also see the responses to your previous discussion.

• ▶ Reply

Permalink Reply by Dean Lamborn on October 29, 2010 at 7:51am

Sounds like a good project for you to add these features.

The wish list for ArduPilot is probably a 10 year programming task if everything was rolled in.

Between control line flying and military UAS is the middle ground found here.

If you had asked anyone 10 years ago if there would be an affordable autopilot with mapping software to set a route they would have laughed at you.

The math/geometry to correctly calculate moment of itertia is beyond this skill level. Instead PID is used to react to control inputs and flight dynamics.

PID settings require the largest amount of work for anyone setting up a new plane. Don't forget that the PID settings that work for you won't be close to your buddy flying the identical plane with a radically different flying style.

• ▶ Reply

Permalink Reply by Colin Johnston on November 18, 2011 at 11:03pm

PID control loop gains are tuned for a specific response for a specific type aircraft but can accommodate some variance while still proving good control and stability. It is very possible that if you change the plant parameters, then you'll have to re-tune the PID gains(Kp,Ki,Kd). If you model your the plant(Heli/Plane, and possibly servos) behavior then you can investigate PID gains that provide sufficient control response for a given input(step,ramp) or for a variety of inputs. Aerodynamic Models of Helicopters are hard to develop, easier for a Plane, but you have to be very knowledgeable in Controls, Dynamic Modeling and Simulation, Aerodynamics and of course Matlab/Simulink. Without a sophisticated computer model of your RC, you are essentially doing this manually by using your brain, vision, and sensitivity on the controls to measure the plants response to varying inputs(flying) with certain gains..This is called experimental flight testing!..Every newly designed plane or helicopter goes through it. Engineers want to understand the relationships with models, parameters and flight performance.

Moment of Inertia, Mass, Stability Derivatives, Control Derivatives, Lift, Drag, Thrust, Weight are just a few parameters in these Aerodynamic models.

• ▶ Reply