SCB #3: Manual Moving Straight - maxct/paparazzi GitHub Wiki

Introduction

This scenario combine autonomous flight (inlcuded wall-avoidance) with manual interruption. Copter passes default phase and flies in arena gathering height in inair phase. It takes about one minute, and after that pilot drives copter to the initial spot and goes 1 meter straight (forward). Then waits for a while and lands the machine.

In future testings with second copter we are going to put it 2 meters (in our arena) away from the first copter and check, how they will interact. First copter will be flying straight and second can be placed on the same height but:

  • in the direction of flying
  • from one side
  • at maximally distant position.

In this scenario we are going to get etalon flight without diturbance, so the initial position is one and and three quarters to the directional walls (in our arena: 1.5m to the sides, 1m and 3m to the back and front walls). That allows us in one future case to put second copter in symmetric position and compare results. Copter is flying around arena all time and we don't expect serious changes in data behavior in case of manual interruption. But some changes take place and it is necessary for comparing.

Noise Detection

Noise is everywhere, and it is present also in received from sensors data. Noise give some information about conditions, where flight took place. If something changes, amplitude and frequency of noise can become stronger or weaker. So, here is pos_z noisy (original) data set and smoothed by filter one (red lines show period of straight flight):

There is an obvious tendency of getting higher, and small oscillations really can be considered as noise. One of the easy ways for revealing noise is just subtracting smoothed data from original noisy. Here it is a plot:

Assumming that we have periodical signal of noise, mean frequency can be roughly found by counting negative and positive zones of signal, mean amplitude - looking for local maximums and minimums (for more details better to use Fourier analysis - maybe next time).

  • mean frequency - 1.7 Hz
  • mean amplitude - 3 cm (varying from 1 to 8 cm)

We can see that our manual interruption has the same noise as it is anywhere else. This is expectable, because copter is just flying alone and it doesn't seriously matter in which way it was done - via remote control or by program. "Having" noise, we can measure it in multi-copter case and check differences and find out, for instance, the safe distance (without influence each to other).

Behavior Visualization

Here we checked the distances to the walls and height-position of the copter. Having these data we can analyze and compare behavior when second copter is added. Most of data, presented on the plots, is smoothed by Gaussian filter because of much noise - we need to see main tendency not unnecessary details.

Z-Position

Theoretically we know that when copter is manually being drived in some direction, it loses height, because of the motors rearrangement. We obtained such result when were evaluating data. This is a plot of z-coordinate values got during the flight:

Front-Back Distances

We were going straight towards front wall. It could be seen at the plot very well. One meter - and we reached the center of the arena (also copter inertially went a little bit further):

Left-Right Distances

Our flight almost doesn't influence on distance to left and right walls, if go parallel to them. So, we have only oscilations between center line of the arena because of autonomous wall-avoidance movements and not exact straightness forward:

Notes

  • after gathering height copter appears to be rotated on about 180 degrees with changed position and it can be impossible (or senseless) go 1 meter forward. So, because we can't controll yaw, the direction can be another
  • it is hard to stay at one point and go exactly one meter, e.g. some inaccuracy occurs.