Just to whet your appetite more for the soon-to-be-released AttoPilot. Here are some of Dean's recent GPS tracks from autonomous flight with a flying wing. Very impressive.

Views: 188


Moderator
Comment by Sgt Ric on October 11, 2008 at 12:48pm
Amazing how it heads for the next waypoint with absolutely no over-compensating, no "seeking".

The Atto seems to hold it's heading like it's on rails!

Good fuzzy logic for situations like the missed waypoint # 4... it didn't do a useless circle or over-react.
It knows that a near miss was good enough when completing such a high angle turn.
Comment by Mike Walker on October 12, 2008 at 10:35am
Very impressive,
What is the diameter of the hold circle?

3D Robotics
Comment by Chris Anderson on October 12, 2008 at 10:41am
As you can see from the map, it's about 200m
Comment by Mike Walker on October 12, 2008 at 9:27pm
Oops, missed that bit of info.
Thanks
Comment by Michael King on October 16, 2008 at 10:48am
looks like an oops on transition from 5 to 6 but looks really good othewise. From 5 to 6 looked like a hard turn. From 1 to 2 looked perfect though.
Comment by Michael King on October 16, 2008 at 10:53am
I love the hold circle function! We had a set of parameters that we had to set to create hold patterns. It was a collection of 8 offsets. You would put in 200,200 to be 200 meters South and 200 meters west of the center coordinate. You would put in all 8 and create the flight path. We found however was if you created a circle to fly and enterd the holding pattern from the wron direction, your aircraft would look really lost in the sky until it picked up the pattern properly so we would just pick a point north and one south and alternate the two so the aircraft could negotiate the pattern without getting lost.
Comment by Dean on October 29, 2008 at 3:42am
Thanks guys, I appreciate the kind comments.

Sgt Ric: It's not "fuzzy logic" per se, rather my own concoction of departure waypoint sequencing. User defines a "satisfy" radius, and once Atto gets within that distance to a waypoint, then it is allowed to abandon the waypoint ONLY when distance to waypoint starts to increment rather than decrement. With the 5th decimal lat and lon (circa 1 meters math resolution) and honed navigation methods of my own creation from first principles, Atto V1.7 routinely runs your plane DIRECTLY over the waypoints. There is actually another "satisfy" radius that is internal to Atto and fixed at 10 meters... once inside this smaller satisfy, the roll target is locked without update. If this is not done, a nasty "wiggle" happens at the last instant when heading error quickly grows huge as the plane veers within mere 1-3 meters of the waypoint. Rather than have this last instant huge heading error make the plane go to max bank angle, I just lock the roll angle target at whatever it was, and let plane cruise through the last 10 meters. As stated above, Atto runs RIGHT OVER most waypoints and it is smooth without wiggle.

Every single waypoint was hit within the user-defined "satisfy" radius of 50 meters. In slower planes you can set it as small as you want, down to 10 meters. If satisfy is too small for the wind conditions and/or flight speed, then plane would circle back to hit the waypoint. AttoPilot NEVER abandons a waypoint, and must check them off in order.
Comment by Dean on October 29, 2008 at 3:57am
Michael: Atto's hold circle has 3 user-configurable parameters:
1) radius (in meters)
2) direction (CW or CCW)
3) altitude.

If plane is done with waypoint path it then approaches the hold circle location at a tangent to the circle no matter the direction it is approaching from, that is if plane is outside the circle. If inside the circle, it will calculate a heading target that is proportional mixture of away from center and a merged tangent, and the mixture of the two depends on haw far plane is between center and the circle edge. So, as plane veers away from center, it gently merges onto the circle as a spline.

Sgt Ric - you commented that Atto holds heading like it is on rails... this is my "proprietary" (can I call it that without a patent?) new navigation methos that I started to call "Tru Trak" before I realized that name is already taken for an autopilot system sold for civilian aviation aircraft (Cessna's, etc). Anyway, mine is a line-holding navigation method I thought up back in June, and hammered out in July. It is really intended for very long runs so you don't waste lots of waypoints to keep the plane on an intended line, such as might be used for photo grid. All you have to do is specify the end points. Atto will veer plane onto the line and can hold it to within 0-3 meters. Because I am using a 32 bit processor and 64 bit overflows in certain places, I have enough trig precision to calculate distane from plane to line with around 0.02% error. This means when plane is up to 10km distance from line, the distane to line is known to at worst 2 meters. Generally our small UAVs will be within a few hundred meters of line, thus the distance is known to 0 or 1 meters. This precision is important to mediate being exactly on this imaginary line even if last and current waypoints are many miles apart. This REALLY comes in handy in a cross wind. Simpler commerical hobby autopilots that merely target the next waypoint and not the line will happily let your plane drift far from the line between waypoints. Also, the user has full control of this Tru Trak navigation in terms of approach cone angle, and a special proportional band width around the line... think of it as a highway road width

Comment

You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service