Chapter 7: Testing the Alternative Navigation System
A number of tests have been conducted on the ANS, from
outdoor testing around the streets of Ottawa, verification
of steps taken to steps recorded, and indoor testing around
the halls of the Mackenzie Building.
The first tests done were to determine whether the ANS
could track a user around a city block. As can be seen
from
Fig. 41
, the system closely tracks the user around the
city block with only slight errors in direction. These
errors can be corrected by increasing the resolution of the
degrees recorded from the compass. Currently the compass
offers one-degree resolution with two-degree accuracy, but
the ANS maps this to ten degrees. The ANS would need an
array of 90 bytes to store this higher resolution and by
doing so would increase the overall accuracy of the system.
Unfortunately, this means changing the microcontroller
again, which is something that should be done when this
system becomes a product, as will be discussed in Chapter 8.
The ability of the system to accurately record the correct
number of steps taken is very important to accurately track
a pedestrian. In
Fig. 42
a test was conducted that shows
that the pedometer accurately recorded the number of steps
taken with an error of 1%. One hundred steps were taken,
50 going northeast, and then another 50 going southwest.
The actual steps recorded were 51 going northeast, and 50
going southwest. Therefore, after 100 actual steps taken,
101 steps were recorded. The numbers at the top of the
screen show the actual number of steps taken in a
particular direction. From left to right, each number
represents the number of steps taken from 0 degrees to 350
degrees, with a step size of 10 degrees between numbers.
Therefore, in
Fig. 42
, the first 10 represents 10 steps
taken at 40 degrees from north, the number 16 beside it
represents 16 steps at 50 degrees from north etc. The
error in direction again is caused by the two-degree
accuracy from the compass mapped to ten degrees by the
microcontroller. Over twenty tests were conducted which
verified the recording of steps taken with the error
ranging between 0% and 2%.
Tests were done to make sure the ANS would work indoors.
It was thought that the walls inside would cause the
compass direction to be inaccurate.
Fig. 43
shows the path
travelled on the fifth floor of the Carleton University
Mackenzie Building, starting at the northwest corner and
travelling clockwise around the quadrant. As can be seen
in the figure, the system is not perfect. A complete
square route was travelled with the ending and starting
points being identical. The route ANS tracked was not
completely square and the starting point and ending point
did not match up. A number of factors may have caused this
result, from human error with uneven step size, to the
compass being affected by secondary magnetic fields
indoors. Either way, the tracking ability this system
provides, is more than adequate for the purpose intended.
As shown in the diagram a spike in direction occurred near
the mechanical room due to the magnetic fields present near
this room.
Bridges and underpasses will cause the Vector 2X compass to
exhibit inaccurate heading information because of the steel
in these structures, which produce magnetic fields. These
magnetic fields will affect the compass and the true
heading of the compass will be skewed. Once the bridge or
underpass has been cleared the compass will again operate
normally and the true heading of the compass will be
restored. For small bridges and underpasses, this error in
direction is acceptable. In
Fig. 44
, a
route
was travelled
which went under a bridge, which contained large steel
beams. As can be seen from that figure, when the compass
was under the overpass the direction obtained was
incorrect, but once clear of the overpass the true
direction of travel was restored.
Distance from A-Flora St. to B-Strathcona St. is 420 meters.
Section A through B is collinear, but the bridge effected
the compass reading.
The tests conducted here shows that the ANS is not perfect
but that it can accurately track a pedestrian in and out of
doors, and it has limited success around bridges. The ANS
as it currently stands has a 2% error of steps taken to
steps recorded, and due to memory constraints has a 10°
error in direction. Assuming city blocks of 100 meters
each, after one block with both a 2% error in steps taken
and a 10° error in direction the ANS will have the
following cumulative error as shown in
Fig. 45
.
A denotes the starting point, and B is the actual location
of the pedestrian after travelling a block (100.0 m).
Point C denotes where the ANS "thinks" the pedestrian is
located.
After one block, travelling 10 degrees off the correct
heading, and having a step recording error of 2% the
cumulative error is 17.7 meters. Comparing this with the
GPS having an accuracy of ±100 meters, the number of blocks
without a GPS signal would have to be greater than five
before the ANS's accuracy would be worse than that of the
GPS system. It is also important to note that this is a
worst-case scenario, and once the microcontroller is
changed and two-degree accuracy in direction is achieved,
the overall accuracy of the ANS will improve dramatically.
This can be seen in
Fig. 46
.
After a block, there is only an error of 4.1 meters, or a
4% error. This is comparable to DGPS accuracy with an
error of ±4 meters. Therefore, the ANS can accurately
track a pedestrian, with the revised microcontroller and
improved accuracy, for a block before its accuracy is worse
than DGPS accuracy.
