AttoPilot updates

Dean Goedde of Attopilot fame has posted development news.

Tri-copter successfully flown with IMU AttoPilot in a stabilized and assisted attitude mode (Co-Pilot type mode). Planning to offer as a commercial product, likely as part of the standard IMU Atto autopilot system.

GPS + IMU INS filter for heading (without magnetometers!) with uncorrected drift of 2.5 degrees per minute under real flight conditions (tested in flying wings), this has been added to overcome the latency of GPS heading. The GPS heading values are de-weighted depending on the average amount of turning (dHead/dt) over the last few seconds. As mentioned above, even if GPS (measurement) input % is dropped to 0%, I find the estimated heading remains accurate to within 5 degrees for up to 2 minutes. With even very minimal GPS measurement input, this method stays well converged AND allows very fast turns (up to 60 degree bank angles!) without heading overshoot. This means your UAV can execute a 90-180 degree snap similar to a pylon racer and continue on the course without over-shooting the path or over-steering. And in case you are wondering, NO: wind does not confuse this filter in any of my testing, and CenTexFlyer can tell you the kind of conditions we were flying the IMU system in last week… worst possible.

Pitch Airspeed / Throttle Altitude method added (old default was pitch for altitude and throttle for airspeed), plus ability to mix it with old method spread proportionally from mix% A to mix% B over altitude error amount C to D. The mixed method is producing truly spectacular results in altitude and airspeed control. Specifically for hing-wing trainer and motor glider afficianados this method allows AttoPilot to work with your preferred airframe. Even for flying wings, the method gives superior control and saves power by not allowing over-speed condition in steep descents like the old method could allow. Developed a new control method that is almost 100% self tuning – this method is the one that controls airspeed via pitch. User only has to supply a single tuning constant.

Enhanced Assisted flight mode with max roll = 60 degrees, pitch = 45. You can really crank the turns and climb out in ARC assisted mode now. AttoPilot has absolutely no problems (IMU convergence) with high attitudes and g-forces for extended durations. Also implemented enhanced rate dampening methods in assisted flight mode that allow use of 60 degree rolls without spinning up the angular momentum of the airframe too much.

Power feed-forward scheduling in turns to better maintain altitude. This automatic correction is applied in autonomous mode to add extra throttle depending on roll angle. This method makes a few basic assumptions but is in fact model-based (as opposed to fudge-factor would be like “hey… maybe add 5% extra power if the roll is > 10 degrees”). It is also easy to turn off if you don’t want it, but I find it adds value so I always use it now.

Motor Start via G-force trigger during autonomous launch. Video coming… This optional feature has been hard-set internally to trigger at 3 G’s total felt force. Intended for small hand-launchable sUAS like plank wings (I might make a REALLY tiny plank wing and call it a Planck wing – nerd joke I just made up)

Automatically Tuned Navigation (and it is wind independent as well). Finally took my notes from last year on this one and implemented it in the code. Works as I had hoped. I have even had IMU autopilot customers think they tuned their navigation perfectly and they are all proud of themself, while in reality they had no control over the tuning and the autopilot did that for them. But I will let them enjoy their euphoria and not spoil it.

Some other heavy-hitter modelling and features are coming as well.

Saving of SET.txt to Internal non-Volatile memory. Now you no longer need to have a SET.txt on the micro SD card in order to fly. There are neat little methods to extract the internally settings to the SD card in case you want to know what is in the Atto. These types of features will also be supported through the GCS.

V1.9 created (has most code advancements from the IMU put back into a thermopile autopilot)

Corrected the Xbee modem compatibility issue in code for the V1.8 control board (no hardware fix is required). There had been a logic level issue, but I found a robust fix in code.

Enhanced ground setup procedure. Fix of some old bugs that could allow servo over-travel during ground test, AND added an automatic cycling test of the camera trigger during this procedure.