|
|
Data
These
are some calculations and experiments I performed to study CESAR's
accuracy and the success of my navigation algorithm.
CESAR
is designed to return to its original course after maneuvering around an
obstacle. These tests show how far off the robot was from the original
course depending on the size of the obstacle.
Click
to View Data
The three main factors that affect the robot’s accuracy in returning to
its original course after maneuvering are as follows: Density, Angle in
relation to obstacle (measures given in the graph refer to degrees away
from being perpendicular to object) and the terrain. The robot’s
accuracy relies on the original measure as to how far the object is when
the robot first encounters it. The robot then proceeds for the amount of
time it would take to go the cosine of the angle it turns (to get around
the obstacle) times the distance from the object. With this method, it is
clear that the original measured distance is crucial to accuracy.
Therefore, only two aspects can inhibit accuracy: speed fluctuations
and/or sonar fault. The robot times itself in order to go a certain
distance. It knows its speed and can therefore go so far in so much time.
The variable here is terrain. Rough terrain will slow the speed causing
the robot to go a shorter distance than called for. If it does not go far
enough, the robot will not be able to return to its original course as
accurately. The second possibility for error is sonar accuracy. My tests
show that density of the obstacle and angle of engagement can both greatly
affect the sonar reading. Although other variables exist such as ambient
sound level and distance from object, these are very minor and only affect
the sonar reading on an extremely small scale. In the field, the sonar
never failed to detect an obstacle. In testing the robot, I found that
under perfect conditions, (paved terrain, high density and perpendicular
to obstacle), the robot gets as close as 1/2 a meter to its original
course. Testing also showed that under unfavorable conditions, (sandy
terrain, large metallic obstacle positioned at 45 degrees), the robot was
5 meters and 20 degrees off its original course. Diverse terrain often
caused large errors in the robot’s final trajectory due to the fact that
it was turning slower on rougher terrain which in turn caused the robot to
steer at undesired angles.
|
About the Project