Page 119 - Concise Encyclopedia of Robotics

FALSE NEGATIVE OR POSITIVE F
Sensors do not always react as intended to stimuli, or percepts, in the
environment. This can occur for a variety of reasons, and is known as a
false negative. Conversely, sensors occasionally produce output when no
legitimate percept is present; this is a false positive.
Consider an infrared (IR) sensor.Suppose it is most sensitive at a wave-
length of 1350 nm (nanometers). False negatives are least likely to occur
for percepts at that wavelength.As the wavelength departs from 1350 nm,
the sensitivity decreases,and the radiation must be more intense to cause the
sensor to produce an output signal. The likelihood of false negatives
increases as the wavelength becomes longer or shorter than 1350 nm. Out-
side a certain range of wavelengths, the sensor is relatively insensitive, and
false negatives are therefore the rule rather than the exception. Whether
the failure to produce output constitutes a false negative depends, how-
ever, on the range of wavelengths that are defined as “legitimate”percepts.
Suppose the sensor in the foregoing example is part of a proximity-
detection device on a mobile robot. A laser on the robot, operating at a
wavelength of 1350 nm, reflects from nearby objects in the work environ-
ment. The reflections are picked up by the sensor, which is covered by an
IR filter that passes radiation easily within the range 1300 to 1400 nm,
but blocks most energy outside that range. If the sensor output exceeds a
certain level, the robot controller is instructed to change direction to
avoid striking a possible obstruction. External sources of IR can cause
false positives. This is most likely to occur if the external IR has a wave-
length near the peak sensitivity region of the sensor/filter, that is, between
1300 and 1400 nm. However, if the external percept is sufficiently intense,
it might cause a false positive even if its wavelength is considerably less
than 1300 nm or greater than 1400 nm.
Robot controllers can be programmed to ignore false negatives or
positives, as long as there is some way to distinguish between them and


Copyright 2003 by The McGraw-Hill Companies, Inc. Click Here for Terms of Use.