SCB #4: Distance Observation - maxct/paparazzi GitHub Wiki
Introduction
With this scenario we investigate the behavior of a single copter, which is located in the center of the arena. For this, we split the experiments in two parts - an idealized part without liftoff and a realistic part with liftoff. The most relevant values we considered for this experiment are the distances to the wall. Because the copter is located in the center of the arena the opposite distances cancel each other out. So the difference of Distance_front & Distance_back and Distance_left and Distance_right should be around zero. By postprocessing the recorded data it is possible to display the measured error in distance.
Future relevance
To know how a copter behaves when an additional copter is used, initially the behavior of an individual copter must be evaluated. These results will be compared with future multicopter experiments to observe the occurring sensor errors of the sonar sensor caused by "cross talk".
Idealized scenario (without liftoff)
Description of executed scenario
For this scenario we placed the copter on a table, which is rotatable around the z-axis. Afterwards we put this table in place at the exact center of the arena. 10 seconds after the experiment started, we slowly rotated the table for 360 degrees. Finally we added some single translational movement in x- and y-direction, still looking toward the starting direction.
Data plots and observated behavior
Rotation using a table
In this plot you can recognize that the opposite pairs of distances are nearly the same. So the difference of each pair should be around 0 and cancel each other out. Also it's noticeable that the pairs switch in their distances when a 90 degree turn was performed. If there is a translation in either x- or y-direction the values of the corresponding pair diverge symmetrical.
Conclusions
The copter behaved nearly exactly like we expected before. If there would be a disturbance source like an other copter you could probably see some diffrences and by subtracting each pair we should be able to measure the occuring sensor error.
Realistic scenario (with liftoff)
Description of executed scenario
This is the more realistic scenario we've done to observate the distances in any direction. Like in the idealized scenario we put the copter in place at the center of arena. After the liftoff we tried to stabilize the copter to stay in the middle of the arena. (And as we expected before, it's really difficult to stabilize the copter in one certain position) Nevertheless the copter didn't moved a lot, so the minimum distance to any wall was approximately 100 cm. Additional we recognized a 90 degree rotate (yaw rotation) we could not compensate because the copter can not perform a yaw rotation to cancel out the drift. The flight was finished after about 60 seconds.
Data plots and observated behavior
distances in all directions
For this experiment it's even more difficult to observe the characteristic behavior in the idealized scenario. This is caused by the unavoidable translational movement. Therefore we split the data in the corresponding pairs of opposite direction.
distances in front-back direction
Compared to the data of the idealized scenario the data is considerably more noisy. But still you can see that the distances were about equal besides the translational movement where the distances diverge nearly symmetrical. Also it's noticeable that the copter performed a rotation around approximatly 180 degree. By locating the crosses of front and back you can easily measure the size of the arena in this certain direction. At the beginning the sum of the distances is about 4m, so the direction was the starting direction towards the shorter 3m wall. During the experiment the sum decreases and ends at about 3m. This shows us that the direction of the copter was now towards a "long" 4m wall.
distances in right-left direction
The data in left-right direction is even more noisy than the front-back data. But also here you can recognize the diverging data at translational movement. Same as in front-back direction there was an approx. 90 degree rotation