Crazyflie, consistent crash

[Kalib-Drift]

Hey everyone, I am currently having an issue with the crazyflie following what seems like the exact flight path and consistently crashing because of this.

I have my room set up with 4 LPS nodes in TWR mode, tracking the crazyflie decently and accurately. when using the cfclient I have the crazyflie set to hold altitude, when telling it to take off from the ground it will rise then swing to the left then forward into the wall, my first thought was that this is an issue with its motors but I get the same result when testing on different crazyflies. it's funny how accurate it is, I have gotten to a point in my troubleshooting where I tell it to take off and then run across the room to catch it in a specific spot.

I think it is trying to hold a specific flight path or position outside my area.
how may I tell it to hover at 0,0,0?

[krichardsson]

The first question that comes to mind is how your system is set up? How are your 4 anchors placed? 4 anchors could probably work but there is no redundancy and a couple more is generally useful.

The second question is if you place the Crazyflie pointing in the correct direction, along the x-axis, before taking off

how may I tell it to hover at 0,0,0?

Check out the [examples](https://github.com/bitcraze/crazyflie-lib-python/tree/master/examples/autonomy)

something like this (not complete code)

    with SyncCrazyflie(uri, cf=Crazyflie(rw_cache='./cache')) as scf:
        with PositionHlCommander(scf) as pc:
            # Hover for 5 seconds
            time.sleep(5)
            # Go to 0, 0, 1
            pc.go_to(0.0, 0.0, 1.0)

Note: you probably don't want to go to z=0 as this is the floor

[Kalib-Drift]

when testing I haven't had any issues with losing anchors while flying, they maintain a green status 100% of the time.

I am measured in meters, and I have set up and re-set up my room multiple times, it is possible that one of those times I made a mistake but unlikely that I made the same mistake every time, the 3d representation in the loco tab seems accurate to my setup and its easy to tell where the crazyflie is in the room.

The position is quite stable when on the ground (a little off in the z value) or on my workstation, when holding it, there seems to be some sway in the tracking when moving quickly but when moving calmy it seems accurate.

I am currently setting up my space with clamps (taping the anchors to the long arm of the clamps) to suspend the anchors away from the pillars and also make sure they will not move at any point.

I got a relatively good flight test at one point while removing one of the anchors from the system but was unable to replicate it, seemed like that anchor was causing undeeded noise.

my hope is that by suspending the anchors away from the metal pillars and cement floor i will decrease the sway in the tracking when moving quickly.

I will look at increasing the anchor's power, alternatively, how might it change its position refresh rate? are they at 100Hz by default?

[krichardsson]

I will look at increasing the anchor's power, alternatively, how might it change its position refresh rate? are they at 100Hz by default?

In TWR the Crazyflie is ranging with one anchor at a time (round robin), I think it is doing one ranging every 4 ms or there about. The distance data is fed into the kalman estimator that runs at 100 Hz. These rates can not be configured.
The graphics in the client is updated at 1 Hz (I think) but if you want a faster update you can use the plotter tab. Set up a log configuration for logs stateEstimate.x, stateEstimate.y and stateEstimate.z

[Kalib-Drift]

after setting up this system again, taking new measurements, and setting up the anchors so they are suspended and secure, it seems to reduce the severity of the issue, but it's still very much present, at this point I'm fairly certain this is simply an issue of the accuracy in the positioning system.

what are some things I can do to increase this accuracy, or what are things that generally cause inaccuracies? What is an acceptable amount of error in the anchor positioning?

does the orientation of the anchors matter?

[krichardsson]

The best placement of the anchors is to keep them "standing", that is with the antenna up and connectors down. The measured distance between two units (anchors or Crazyflies) changes quite a lot (probalby around 40 cm) depending on the relative orientation of the antennas. The antenna radiation pattern is fairly uniform when the anchors are standing up and this is why this is better. This is generally a pretty tricky problem to solve, you can read more about it here if you are interested.

The environment around the anchors (walls, objects, humans) also effect the measured distance and the estimated position. The problem is that it is very hard to predict what kind of effects you will get, I think trial and error is the only way I can recommend.

[Kalib-Drift]

Alright, I have accounted for most of these issues but I will try setting up the space with upright anchors.

I have already sent an email to your team but it would be great to set up a meeting to potentially go over this issue in a more efficient way.