Arducopter 2.0.38 Flight test - Wind effect - Blogs

I want to add this discussion to record experiences in flight test from 2.0.38 and on mainly in RTL, LOITER and AUTO. Also to analise the effect of the WIND, that for me is THE big problem. I´ve tested several days my copter in different wind conditions. Below is a summary of my conclussions.

RTL flight tests

We can say that RTL works, but depending on wind conditions and direction of the wind in the moment of RTL.

Upwind. This is the best situation. The copter aproaches slowly the goal in a straight way more or less ( the goal is closed to the cross of roads in the images)

wind = 10 knots SSW

SUCCESFULL TESTS

(log 26_file 1 attached)

The example below, ended well but it makes a path not logical to reach the goal (why?)

(log 21 file 1)

Below is other succesfull RTL, but the path took one complete turn to reach the goal probably due to the wind

(log 23 file 1)

NON SUCCESFULL TESTS

In all this test the copter diverged strongly and it was recovered THANKS to the SIMPLE mode, that is essential to survive to big crashes.

Below is a tipical divergence. It starts downwind and then enter in a divergent spiral. (Red curve shows the final SIMPLE mode to save the copter).

Downwind - other diverging RTL

(log 28 file 1)

Other problem that happens to me is the delay in reacting once the SIMPLE mode is activated to save the situation. The yellow curve shows when SIMPLE is activated but it responded some seconds later.

(log 22 file 2)

Upwind RTL. Another example of bad path to reach the goal

(log 19 file 2)

Tomorrow I will add more LOITER and AUTO (exciting) experiences.

You need to be a member of diydrones to add comments!

Join diydrones

Comments

afernan August 7, 2011 at 1:25am

AUTO flight tests

Well... this is really exciting.

First to comment: the function TAKEOFF. When activated AUTO the copter lift up to the starts and beyond... I repeated several times (4-5) and only in one case (just after restarting and wayting for GPS) it did what was expected (lift up to 6m and wait some seconds) . So I decided to eliminate TAKEOFF and simply use a first WAYPOINT close to home at 6m heigth. It was not recorded in any log (I don´t know why)

This is the goal path:

This is the first try. Started from HOME in the floor. Then when to WP#1 but due to the wind effect, a semicircle path is done. Then overshoots at WP1 and started a diverging spiral. Then, I recovered the copter in SIMPLE mode as usual (purple path) (log 11-12 file 1)

Wind is allways from SSW - about 10-12knt

Second try, exactly the same process.(log 14-15 file 1)

Wind increased a bit. Below shows two tries together (red and blue line) that diverged before reaching WP#1. Purple line is in SIMPLE mode to recover.

In conclussion:, something to do with the wind perturbation.

This ends the first set of flight tests reports.

Hope it will serve to improve the code, and animate others to report their experiences.
afernan August 6, 2011 at 12:26pm

Jason:

going upwind it works most of the time, but downwind overshoots also home location (helped by the wind) and then start to diverge (see figure of case log 28- file1) . I agree that a 2D rate controller is the solution.

In general, if the copter goes out of the line of the wind direction (either in loiter or RTL or WP) then there is a big probability of entering in a diverging spiral (that acelerates more and more). Sometimes a simple gust, while in loiter or RTL, initiate the diverging spiral.

Angel
Developer

Jason Short August 6, 2011 at 11:57am

Just got back from tests and I think I know the problem now. Basically the crosstrack correction is making the quad go too fast and overshooting the home location. That's why downwind and upwind work and crosswinds overshoot. I basically have a 1 dimensional rate controller and I'm going to have to change it to a 2D rate controller. In the mean-time, turn Crosstrack kP down to 0 and see if that helps overshoot.

Jason
Developer

Jason Short August 6, 2011 at 10:55am

I am doing tests today. I'm going to make some changes based on this and go fly them right now.

Thanks!

Jason
Moderator

Roberto Navoni August 6, 2011 at 7:36am

Hi jason,

i doing the same test , and i have same result of your . I saw that when there is a windy day with rtl the quad doing a semi cirlce when going to starting point.

Best

Roberto
agmatthews August 6, 2011 at 1:33am

My understanding of the nav update code in ArduCoptermega.pde is that the RTL mode switches to Loiter mode when close to the Home location

case RTL:
// are we Traversing or Loitering?
wp_control = (wp_distance < 4 ) ? LOITER_MODE : WP_MODE;

I assume that the '4' in the code is meters

Looking at your examples it seems that the issue is in the Loiter mode as the quad copter nears the home location.

Have a look at the 2.0.39 Code. It has a new set of PID values for Loiter

It may help

-Andrew
afernan August 6, 2011 at 1:09am

continue reporting...

LOITER flight tests

In general, it works. But sometimes it diverge due to the wind (probably) and PID settings.

Below is a typical LOITER with light wind (8-10 knt). It remains in about 4m radius circle.

(log 42 file 1)

Other succesfull example, same wind conditions. Smaller radius (3-4 m)

(log 8 file 2)

Below is an example of starting well, but after a while, diverging (similar than loiter divergency) due to wind gusts increase, probably

(log 13 file 2)

Below is a loiter with more wind (12knt), still stable but more radius (10m)

(log 39 file 2)

AUTO tests reports coming soon...
afernan August 6, 2011 at 12:56am

Not yet, For the moment I´m trying with defaults and learning how it works.
3D Robotics

Chris Anderson August 5, 2011 at 8:09pm

Nicely documented! Have you tried turning up the Nav P or I terms?

This reply was deleted.