Filtering poses going into the controller

[PrajwalBBharadwaj]

We want to filter out bad pose values going to the flight controller.
Below is a part of the code we are using to get the quadrotor to takeoff and land.

`

def run_demo(self):

    try:

        swarm = Crazyswarm() 

        timeHelper = swarm.timeHelper

        timeHelper.sleep(10)  

        cf = swarm.allcfs.crazyflies[0]

        cf.takeoff(targetHeight=1.0, duration=self.TAKEOFF_DURATION)

        timeHelper.sleep(self.TAKEOFF_DURATION + self.HOVER_DURATION)

        cf.land(targetHeight=0.04, duration=2.5)

        timeHelper.sleep(self.TAKEOFF_DURATION)

        return True

    except Exception as e:

        print(e)

        return False

`
We see in the code that cf.takeoff and cf.land are services that are executed by the server-client pair. I am assuming that the flight controller is somehow accessed by the functions. The controller has to be a closed loop and hence, should have continuous access to the pose data. But we aren't really providing that data through our code here, so I am assuming that the flight controller somehow internally has access to this pose data. Since I want to filter out bad pose values, I would like to know how the controller is getting the instantaneous poses data on which it is acting and getting the quadrotor to the goal. Thank you!

[whoenig]

The poses a provided by motioncapture_ros, which publishes a /poses topic. The crazyflie_server node subscribes to it and sends it the Crazyflies via broadcasts. This data is fused onboard using an extended Kalman filter to keep track of a state estimate. The controller also runs onboard, comparing the estimated state to the desired state.

Options for filtering:

• Use librigidbodytracker rather than your motion capture software, which has a simple rejection filter.
• Increase the standard deviation of the motion capture data in the Kalman filter
• Remap the /poses topic and write a node that has your custom filter.