Page 1 of 1
Take Off Drift Issue
Posted: Fri Oct 08, 2021 2:16 am
by Barry
Dear all,
I have made a new drone using the bolt. While when I run the example: flowsequenceSync.py.
During the takeoff, the drone drifts too much. But when I try to decrease the flying time during the takeoff, the drift issue will be mitigated. It looks like the drone will take off much faster than before. When I use the Xbox controller to control the drone using the hover mode, the drone could stable hover without drift. My question is:
How can I solve the issue to run the example I mean slowly take off without too much drift?
Origonal code:
Code: Select all
for y in range(10):
cf.commander.send_hover_setpoint(0, 0, 0, y / 25)
time.sleep(0.1)
Modified one:
Code: Select all
for y in range(4):
cf.commander.send_hover_setpoint(0, 0, 0, y / 10)
time.sleep(0.1)
Could someone help me, I would appreciate your answer.
Cheers!
Re: Take Off Drift Issue
Posted: Tue Oct 12, 2021 7:50 am
by tobias
Can you let us know your setup? My guess is that you are using a flow v2 deck and no absolute positioning system?
One issue with the flow deck is that it is kind of "blind" below 8 cm from the ground so the takeoff is "open loop", until the Flow deck starts to get flow measurements. Also close to the ground you get high turbulence which doesn't help the situation. This effect can be improved a bit by having a well balanced and tuned drone. Thus if the drone is flown manually and only thrust is applied when taking off it should go straight up.
Re: Take Off Drift Issue
Posted: Tue Oct 12, 2021 8:26 am
by Barry
Thanks tobias,
I was only equipped with the flow deck on the crazyflie bolt. I have made some modifications to the flowdeck_v1v2.c and zranger2.c
In flowdeck_v1v2.c, I set useAdaptiveStd = true and set OULIER_LIMIT 200
In zranger2.c, I set:
Code: Select all
static const float expPointA = 6.0f;
static const float expStdA = 0.0060f; // STD at elevation expPointA [m]
static const float expPointB = 10.0f;
static const float expStdB = 0.5f; // STD at elevation expPointB [m]
The drone is now better than before but there sometimes appears the drift issue in my room. I used some gray cushions on the ground. Do you have any ideas on solving the drift issue except I put some colorful ground patterns?
You said that below 8cm, the flow deck is blind, but as for the crazyflie. I have one crazyflie in my lab. The flow deck we mount on the crazyflie is actually super close to the ground right. But it can still fly very stable. So the question is: for the larger drone I made using the bolt, should I make some higher landing skids to prevent the drift issue?
The drone setup is shown as below:
Re: Take Off Drift Issue
Posted: Wed Oct 13, 2021 9:02 am
by tobias
Do you have any ideas on solving the drift issue except I put some colorful ground patterns?
More features on the floor will probably improve the situation. This should also be true when it is close to the floor.
So the question is: for the larger drone I made using the bolt, should I make some higher landing skids to prevent the drift issue?
It is worth a try. As it is now it looks very close to the ground. The code could also be improved I think. Right now I think we use the flow values even though it is close to the ground. Possibly ignoring values if they are below e.g. 5 cm would improve things.
Re: Take Off Drift Issue
Posted: Wed Oct 13, 2021 10:59 am
by Barry
Right now I think we use the flow values even though it is close to the ground. Possibly ignoring values if they are below e.g. 5 cm would improve things.
Thanks, tobias. So how can I modify the code to ignoring values below some heights?
Re: Take Off Drift Issue
Posted: Fri Oct 15, 2021 9:25 am
by tobias
I suggest doing it in the measurement model for flow, mm_flow.c. You could try and simply
return from the function if height is below 0.1m.
Code: Select all
// Saturate elevation in prediction and correction to avoid singularities
if ( this->S[KC_STATE_Z] < 0.1f ) {
return;
} else {
z_g = this->S[KC_STATE_Z];
}
this->S[KC_STATE_Z] is the height estimate.