G. Benet, F. Blanes, J.E. Simò, P.
Perez
IEEE Transactions on Industrial
Electronics, Vol.56.NO 10, October 2009
Summary
The paper illustrates a precise and good
method for infrared sensor technology applied on a mobile robot.
IR sensors are generally
used for proximity sensors and for obstacle avoidance in robotics, they have
the great advantage of costing less than ultrasonic sensors and having a better
response time, but of course they have drawbacks such as having a non-linear
behavior and depending on the reflectance of the surroundings, making this a
poor quality sensor in mapping applications. Although these negative points in
IR sensors, still their application is convenient, specially when considering
the long “Time to Flight” in US
sensors and a better angular resolution. Due to these reason, US and IR sensors
can be integrated to form one sensor system, obtaining a precise and accurate
result.
·
Model
The authors have performed
the experiment with a YAIR robot, with 16 IR sensors (two for each side of the
octagon on the belt of the robot) and a US sensor. Each IR sensor, combination
of an emitter and a receiver, is capable of measure distance targets at up to 1
m.
The sensor’s output can be
measured by using the photometry inverse square law, where it’s depending from
the distance of the target and the incidence angle. α and
β are two
parameters which also influence the output of the IR sensor, the first includes
the radial intensity of the IR emitters, the gain of the amplifier and the
reflectivity coefficient of the target, the second is related to the amplifier
offset plus ambient light effect and it can be taken without IR emission, that
is why it’s not in our interest and we can just consider the cleaned signal ‘y’. α is a bit more difficult to estimate and is divided
in α0 and αi, the
first is a constant component and the latter is dimensionless and must be
calculated (as illustrated in the paper).
An interesting variable
which is required to know to understand how precise the instrumentation is, is
the angle of incidence, which can is calculated through iteration, first
assuming the angle as 0 for 2 IR sensors in order to calculate the distance,
later using the calculated ratio of these in order to obtain the incidence
angle.
In operation the sensor
may occur in different types of errors: measurement error, angle uncertainty,
error caused by noise in the reading and error due to wrong incidence
estimation (where the first two can be omitted when considering the overall
standard deviation since they are likely to be treated as distributed as a
zero-mean Gaussian).
Noise in the measurement
can be actually afflicted by a certain amount of uncertainty and it is
demonstrated that it depends on the error component caused by the noise in the
reading of the error of the clean out put.
Parameters αi can actually be estimated and the proposed model
allows us to obtain a table with its values, although is also is affected by
uncertainty.
Key Concepts
IR and US sensors applied
in one system to obtain better precision.
Key Results
The proposed system
permits to obtain αi values
for each kind of surface, demonstrating though that for a certain range of
colors there is no significant change, but being still the only fundamental
parameter of the model.
The sensor presents also
the advantage of providing also the expected uncertainty, as shown through
calculation in the mentioned paper.
No comments:
Post a Comment