DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

KML illustrates cornering performance vs. speed

Prismatic

Well-Known Member
Joined
Feb 26, 2018
Messages
587
Reactions
644
Age
70
Location
Northern Colorado
Mavic Air pilots, especially those who use Litchi, may find this KML file interesting. It shows how a Mavic Air performed on a course specifically set up to see if—and how far—the drone “overshoots” corners at various speeds.

Here's a snapshot of the course, marked for the direction of travel.
View media item 2638 I highly recommend using Google Earth Pro over any browser- or mobile device-based edition of GE.

I used Virtual Litchi Manager to plan the mission and export it to Google Earth. (If you’re unfamiliar with VLM, you owe it to yourself to look into it.) The following KML members come from VLM output:
  • Waypoint Path (skeleton) : renamed component (Yellow path)
  • Expected Track (smoothed WpP) : renamed “Smooth Flight Path” (Black path)
  • Diagnostics folder and contents : standard components
After I actually flew the mission—three times at two different speeds—I imported the actual tracks from each flight into Google Earth, using KML files that I’d exported from Airdata. The following KML members come from Airdata:
  • 20190121-8mph : Green path (13kmh)
  • 20190121-18mph-1 : Red path (29kmh)
  • 20190121-18mph-2 : Purple path (29kmh)
  • Thus, the KML shows not only the desired flight path on this Waypoint Mission, but three instances of the actual, in the field, flight path.
About the Mission

The drone flies a series of right turns: first 30°, then 60°, then 90°, then 135°. Each turn employs a 20 ft smoothing curve and has a long, straight approach of ±160 ft (±49m).
Then it runs a quick Z-curve before executing a "tilted pseudo-orbit“ around a landmark. The orbit is defined by a series of equally spaced and equally smoothed Waypoints, and has a radius of about 56 ft (17m). It finishes with a 180° “full reverse” maneuver before coming home.

About the Flights

All three flights were taken in mild to moderate winds:
  • 8mph flight : SW @ 3.0mph, gusting to 5.6mph (1.3m/s, 2.5m/s)
  • 18mph flight 1 : NW @ 6mph, gusting to 9.5mph (2.7m/s, 4.3m/s)
  • 18mph flight 2 : SW @ 5.3mph, gusting to 7.1mph, (2.4m/s, 3.2m/s)
The flight paths are interesting. Not surprisingly, the Mavic Air tracks closer to the course—and recovers from deviations faster—at low speed than it does at speeds near the drone’s performance limit.

For example, the low-speed track (in green) is nearly true to the course, except in the final hairpin turn, and in the Z-curved approach to the orbit, where its maximum deviation is about 3.3ft (1m). High speeds tell a different story.

The first high-speed flight (in red) met the strongest winds near the point on the ‘orbit’ where the track shows a massive deviation of 18⅔ft (5.7m) off the course. Not only did the drone have to fight a 9mph headwind, the course also gains altitude at that point. Still, I'm a little surprised that the Mavic Air didn't have the reserves to better execute the maneuver; that’s a lot of drift! Also, I can’t explain the discontinuity opposite the point of maximum deviation in this flight. My best guess is a wind gust, but holy smokes, at 18mph it should take impact from an inbound pass to cause that kind of lurch. There are no logged events to suggest anything.

Most surprising to me was that all flights held true on the 135° corner, whereas they overran the course on lesser curves.

That’s particularly true, because I ran a prototype of this course the day before, at 17mph. That flight isn’t in this KML because parts of the course differed then, but the 30-60-90-135 section was nearly identical. In that preliminary test, the drone massively overshot the 135° turn, and took nearly the whole distance to the Z-curve to return to the course. (Even more dramatic than the deviation shown in the KML coming off the final hairpin turn in the red path.) The only difference was that in that version, the 135° turn had a 40 ft. smoothing curve, rather than 20 ft in this KML. The fact that requiring a sharper maneuver elicited far better performance is utterly baffling. Similarly, I expected a notable overrun on the “full reverse” maneuver near the end. Again, the drone seems to have “turned on a dime”, without overshooting the reverse-point in any significant degree.

Obviously I’m missing something, though the results are largely in line with what I’d expect.
 
Last edited:
  • Like
Reactions: rfc
This is a very interesting and revealing experiment.

CW vs. CCW are very different physical events depending on geographic latitude. Additionally, every airborne craft has a favorite way to turn, the opposite always more difficult to execute to the same degree of fluid accuracy. Thus you have at least two key influencing factors.

Fixed and rotary wing pilots are typically aware of the increased challenge making turns in the least favoured direction, and wind can exert a dramatic difference in aircraft response while turning this way, often in unpredictable fashion. Thus pilots need to be on their game to execute accurate maneuvers in both directions, one of the key challenges in acrobatic flight.

Keep in mind the MA programming probably isn’t as concerned about getting back on course as it is arriving at the next waypoint coordinates.

As for the other deviations you’ll have to continue your investigation, perhaps walking the course with a magnetometer would prove interesting.
 
Last edited:
  • Like
Reactions: Prismatic
CW vs. CCW are very different physical events depending on geographic latitude. Additionally, every airborne craft has a favorite way to turn, the opposite always more difficult to execute to the same degree of fluid accuracy. Thus you have at least two key influencing factors.
Is the CW/CCW difference on account of forces applied by the engine(s) to the airframe? Your reference to latitude is puzzling; surely Coriolis effects are too negligible to notice at this scale. Or not?

Now, I can imagine that such engine-originated forces could result in a favored turn direction, but multi-rotor craft have opposing rotations in the engines, which I would think cancel out. No?
Fixed and rotary wing pilots are typically aware of the increased challenge making turns in the least favoured direction, and wind can exert a dramatic difference in aircraft response while turning this way, often in unpredictable fashion.
Perhaps I should reverse the course and see how that plays out. Sadly, I haven't found an easy way to reverse a mission. But, yeah, when you're pushing your aircraft to its performance limits, its ability to "stay the course" suffers the most in wind and/or altitude changes, as the red track illustrates.
Keep in mind the MA programming probably isn’t as concerned about getting back on course as it is arriving at the next waypoint coordinates.
I'm not so sure; you'll note that in most cases you can clearly see the drone returns to the course long before it arrives at the next WP. But I have no deep insight into that question; perhaps others (@sar104? @msinger? Anyone?) do.
As for the other deviations you’ll have to continue your investigation, perhaps walking the course with a magnetometer would prove interesting.
That location was once a shallow agricultural reservoir. It was backfilled by the City to create a neighborhood park, and the source of that backfill is unknown; it may or may not contain concrete chunks with rebar. That said, I've never had the slightest compass-related issues there, not even a request for compass calibration. That one lurch in the red path remains a puzzle, but if it was magnetically induced, why on one flight only? Anyway, magnetic interference isn't at the top of my list of questions.

Thanks for your thoughts! I'm not a manned-vehicle pilot, and that "favoured" turn direction business is news to me though I can't say I fully understand the reasons. This test mission has only right-turns (one exception), which may have biased the outcome in some degree.
 
This is a very interesting and revealing experiment.

CW vs. CCW are very different physical events depending on geographic latitude. Additionally, every airborne craft has a favorite way to turn, the opposite always more difficult to execute to the same degree of fluid accuracy. Thus you have at least two key influencing factors.

Fixed and rotary wing pilots are typically aware of the increased challenge making turns in the least favoured direction, and wind can exert a dramatic difference in aircraft response while turning this way, often in unpredictable fashion. Thus pilots need to be on their game to execute accurate maneuvers in both directions, one of the key challenges in acrobatic flight.

Keep in mind the MA programming probably isn’t as concerned about getting back on course as it is arriving at the next waypoint coordinates.

As for the other deviations you’ll have to continue your investigation, perhaps walking the course with a magnetometer would prove interesting.

These quadcopters have net zero angular momentum and so they are completely symmetric in their turn response.
 
  • Like
Reactions: Prismatic
Corilis has an effect, not to be discounted, even the KAMAN K Max will rotate without stick correction. In multi-rotor aircraft, in this case quads, counter-rotation does not cancel out because the motors are continuously operating at different rpms to constantly maintain the flight path managed largely by the onboard flight controller utilizing onboard sensor input, as directed by the pilot from the RF controller. Counter-rotating motors correct many problems but do not ‘balance out’, in many flight conditions not even close. Even though the design concept is for ‘zero angular momentum’ it is impossible to eliminate in every flight orientation and atmospheric condition.

Reversing the outdoor mission cannot replicate the identical conditions of the original flight, and although it might yield some interesting data, likely multiple missions would need to be flown to extract reliable information, best undertaken in a zero wind location.
 
Corilis has an effect, not to be discounted, even the KAMAN K Max will rotate without stick correction. In multi-rotor aircraft, in this case quads, counter-rotation does not cancel out because the motors are continuously operating at different rpms to constantly maintain the flight path managed largely by the onboard flight controller utilizing onboard sensor input, as directed by the pilot from the RF controller. Counter-rotating motors correct many problems but do not ‘balance out’, in many flight conditions not even close. Even though the design concept is for ‘zero angular momentum’ it is impossible to eliminate in every flight orientation and atmospheric condition.

Reversing the outdoor mission cannot replicate the identical conditions of the original flight, and although it might yield some interesting data, likely multiple missions would need to be flown to extract reliable information, best undertaken in a zero wind location.

It doesn't matter that the angular momentum is changing during a turn - it has to do that. What matters is that the system is entirely symmetric, i.e. the net angular momentum changes are equal and opposite for CW turns vs. CCW turns in still air. Now if you are talking about turns into the wind being different from turns with the wind, then that's a different thing altogether - it's an external factor. But there is no intrinsic difference for a CW vs. CCW quad turn. Mechanically, a mirror image of a quad is indistinguishable from the original.
 
  • Like
Reactions: Prismatic
Yes, that's the science part of the equation.

If that's a response to my last post then I have no idea what it is supposed to mean. Let me ask a more specific question. If you think that a CW turn is different in any way to a CCW turn, then which relevant variables or derivatives can you point to that would would not be identical in magnitude and opposite in sign between those two cases? In other words, how, specifically, would a quadcopter turning CCW differ from a mirror image of a quadcopter turning CW?
 
  • Like
Reactions: Prismatic
These quadcopters have net zero angular momentum and so they are completely symmetric in their turn response.
That was my intuition. Thanks for responding. I tagged you because you're about the most knowledgeable person I know of regarding the flight characteristics and internal affairs of drones like this.
 
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Forum statistics

Threads
130,586
Messages
1,554,115
Members
159,586
Latest member
DoubleBarS