Bitcraze Forums

Hi ,I find the latest version TDOA2 will often appear sudden impact crash, I'm not sure whether it is my Quadcopters' problem or others, but I have test 10 Quadcopters and found crash more easily than TWR, I don't know what cause it ? Maybe the firmware have influence on it ? can you give me some help ?

Hi,
When the kalman filter is fed very bad data we have observed crash. Until recently TDoA2 had no outlier rejection, which causes these crash, we have mostly observed it when starting the CF2s outside the system convex hull.

We recently added a minimal outlier detection: we reject TDoA measurements that are higher than the distance between the two involved anchors. This is a pretty fundamental check and it seems to help a lot the EKF crash.

Can you try the latest master branch of the firmware to see if it improves the situation for you too?

Thank you for your answer. I have try the latest master branch of the firmware but it also happened when I control to flight in a circle.Maybe my Quadcopters are not stable enough or the position estimation of kalman filter isn't faster enough to run a quickly circle.

Sorry, I think I missunderstood the problem.

Are you observing firmware crash, or real physical crash?

If it is real physical crash, the problem is much different .

Theoretically we would expect TDoA to be a bit worse than TWR. However we have observed very good performance in our lab that shows very similar performance.

One big difference between TWR and TDoA is the importance of the anchor placement: TDoA is working best within the anchor convex hull, if you fly outside the convex hull the accuracy will decrease very quickly.

How is your system setup, ie. do you have the Crazyflie in the anchor convex hull?

Thank you for your help.I observed a real physical crash.I setup my system in a rectangle and fly inside the convex hull. I think it can be my quadcopters' physical problem. Now I try to repair them.Do you have some suggestion to test a quadcopter is enough stable to fly ? It often performed very well in last time but performs a sudden crash in this time.

You usually can hear when a Crazyflie will fly well, the flight performance is mainly affected by vibration and vibrations will make the quad more noisy. Too much vibration corrupts the accelerometer measurements. One of the main source of vibration is unbalanced propeller. It is possible to balance the propellers to optimise performance: https://www.bitcraze.io/balancing-propellers/. Vibration can also come from the motors axis being bent, it can happen after many bad crash, in that case changing the problematic motor will help. There is parameters in the Crazyflie to force a PWM values to individuals motors, this can be used to test each motor/propeller individually.

Though, I am still curious about "sudden crash". Vibration will degrade performances but 'sudden crash' sounds much more "binary". Are you flying very aggressively at the edge of what the Crazyflie is capable and so performance degradation can cause crash, or are you observing sudden crash while the Crazyflie is stationary or moving slowly?

Hi,
I also observed sudden physical crash while the Crazyflie moving slowly. I was confused about this.And when I repaired it and it can also perform well. I couldn't dope out any solution about this problem.

Do you mean that it is now working well after repair?

If not, would it be possible for you to log the position estimate and take a video when a 'sudden' crash occurs. This is not somethings we have observed so I am curious to understand more what could be happening.