Posted by Koen Delvaux on December 27, 2008 at 9:06am
Today my autopilot hardware took the sky for the first time. Although this maiden flight was successful, its only a first step.
The primary goal of the test was to see if the autopilot failsafe mechanisms operated fine:- no lockup on division by zero- instant regain of control after brown-out causing reboot- no lock on uncaught interrupts- handover to servo failsafe positions on RC reception lost- instant regain of control when RC signal returns- no erroneous lockups in main state machine when rebooting whilst in flightAll mechanisms operated fine and none where triggered during normal flight. This proves the autopilot is a stable and reliable platform for airborne operation.Built in mixer test.The autopilot has a built in mixer. On first use you need to tell the autopilot what the channel and mixer arrangement of your transmitter/airplane combination is. To put the autopilot in calibration mode, switch on the transmitter, put all sticks in the center and switch channel 6 to "ON". Now turn on the airplane.At this point you can calibrate the autopilot by putting all sticks in the center and subsequently actuating the aileron stick from full left to full right and back to center, followed by elevator (fullup, then down, then center), rudder (left, right, center) and throttle (zero throttle, max throttle, center). The autopilot will confirm completion of the procedure by turning the rudder full left, then full right and then back to center. From this moment on, you can operate the airplane with the controls as normal, but the calibration values are not saved yet.The last step in calibration is to calibrate the IR sensor for horizon detection. You do this by holding the plane level (or even better: flying it, turn down the throttle and trim for a best glide scope) and switching channel 6 to "OFF". At this moment, calibration values for the transmitter arrangement and the IR sensor are saved to the permanent memory of the autopilot.When you now fly the plane and engage autopilot mode, the autopilot uses the same channel mix,offset and endpoint adjustment as you programmed in your transmitter.This all works fine. When you switch to manual control, channels are passed transparently. This also works fine.Only when using part automatic / part manual control (like when using the autopilot for stabilization alone and operating rudder/throttle manually), the manual controls don't have the same midpoint and endpoint settings than when in manual mode. This still needs to be looked into.I would like some feedback on the first-use calibration procedure. Do you think it is valuable to have support for random channel assignments/mixes/atv/trims? Do you feel the initial calibration procedure it too complicated?
Read more…
Posted by Koen Delvaux on December 21, 2008 at 4:59am
I have been working on/off on an autopilot to use in RC airplanes. The goal is to fully automatically take pictures of an object like a house, construction site, etc.I made the hardware in last year's Christmas holiday. During the year I spend a few hours writing the embedded software for it (in C, works much faster then assembly).The hardware is super low cost: The PIC dsPIC30f4012 is very cheap and still has enough memory and speed (30 MIPS) to be self sufficient autopilot.
The goal for this Christmas holiday is to do in flight testing of the hardware and stabilization routines.
Read more…
