‹date/tim=
e›
‹footer›
‹#›
------=_NextPart_01C9D49C.122AA0A0
Content-Location: file:///C:/0B665CD1/GE423website1_files/master83.xml
Content-Transfer-Encoding: quoted-printable
Content-Type: text/xml; charset="utf-8"
![](3D"master83_background.gif")
![](3D"master83_image001.gif")
HOME
GE 423
Spring 2009
=
Group 3
![](3D"master82_background.gif")
![](3D"master82_image002.gif")
Colleagues
►Kyle Hewerdine <=
/span>- =
Me=
chanical
Engineering
=
►Eric Langendorf –
► General Engineering
►
►Billy Poetker –
<=
/span>
`=
58; General Engineering
►
►Joe Zearing –
`=
58; Mechanical Engineering
![](3D"slide0002_image004.jpg")
Strategy
►The group strategy for the fin=
al
project was to figure out which sensors behaved the most
accurately and consistently. After which, we
utilized those sensors to carry out the objectives
stated. Consistency was the key for our de=
sign. Once
the robot performed consistently, the rate at wh=
ich
the robot moves through the course was ramped up to=
find
the optimal operating velocity. Our robot collects =
go=
lf
balls while simultaneously carrying the can to its drop off
destination. The robot used the LADAR as a =
re=
placement
for a lot of the IR and Ultrasonic functions or=
iginally
used. The VB application was used to track the ro=
bot’s
progress through the course, as well as being able to
change variables mid-run. This allowed for more =
ac=
curate
tuning and debugging of the robot.
LADAR
`=
58;
►
Our group used LADAR in several ways including with a Kalman filter=
to
update the dead reckoning position, to determine when to dr=
op
the can off, and to determine when to stop moving forward at the first
light. We also coded, but never used several
emergency relocation conditions that would be triggered if the LADAR were =
to
get lost. These algorithms proved unnecessary
because the LADAR kept an accurate position during the entire contest.
=
;
►
The main use of the LADAR was to fix the dead reckoning
drift. This was accomplished using the code that =
=
was
provided to us. In this code, the LADAR data was converted from
local coordinates to global x-y positions. Then <=
span
style=3D'position:absolute;top:29.5%;left:8.05%;width:99.25%;height:3.0%'>=
th=
e data
was sorted through to identify which points represented
corners. This was done by checking the slope be=
tween
adjacent points. If a corner was identified within a threshold
radius of where the dead reckoning said a co=
rner
would be, the LADAR code linked these corners together. After
sorting through all the data, if there were at le=
ast
two corners linked together, a translation and rotation were generated to
correct the position. This correction wa=
s then
put through a Kalman filter to update the dead reckoned
position. The Kalman filter works by assigning un=
certainty
values to each source of information and then applying these uncertainties=
to
produce a best guess of the current position. In
this way, many sources of information can be combined to produce a low
uncertainty measurement.
►
The LADAR was also used in our can drop algorithm. The r=
obot
would arrive several feet in front of the light and
pause to see where the light was. If it did not see the light, =
it
would turn until the camera identified a sufficiently la=
rge
blob in its field of view. The coordinates of the blob in the
camera frame were used to direct the robot towards the
light. Once on its way, the LADAR distance reading from the fro=
nt
center was used as a threshold to determine wh=
en we
were close enough to the wall to drop the can. Occasionally this
method would not work well because the LA=
DAR
reading would represent the distance to the servo post attached to the lig=
ht
or the plastic light guard instead of the wall behind these
items. If this happened, the robot would drop the can too early=
and
so another method of dropping the can was
created. This method also used the LADAR data, but in a slightly
less direct manner. We simply told the robot to drop =
the
can once it reached a certain y position. This only worked well
because the LADAR ensured that our position was =
very
accurate.
►
The third use of the LADAR was to determine when we were close enou=
gh
to the first light to have tripped the overhanging IR
sensor. This just used the center LADAR reading in a threshold =
to
move us to the next step.
►
A few more uses for the LADAR were coded but not implemented because
they were deemed unnecessary. These
include using the northwest and northeast corners as absolute dead reckoni=
ng
correction points. We would have used the LADAR read=
ing
for the left, center, and right distances to correct our absolute <=
/span>po=
sition
if we had been lost. To arrive at these corners, we told the ro=
bot
to turn to the north via its compass, and th=
en
drive straight until it saw a wall. Once at this wall, it would
wall follow until it either saw that it was in the no=
rthwest
or northeast corners marked with green, or it would be following the middle
wall section. If it reached the en=
d of
the middle wall section without seeing the green, it would turn north again
and follow until it knew where it was. Another approa=
ch
that we coded but did not end up implementing was to triangulate the posit=
ion
using the two colored corners. Th=
is
would have used the LADAR to check the distances calculated using the came=
ra
data centroids.
Pictures
![](3D"slide0005_image006.jpg")
![](3D"slide0005_image008.jpg")
![](3D"slide0005_image010.jpg")
![](3D"slide0005_image012.jpg")