Environmental Engineering Reference
In-Depth Information
In some applications, the sensor is integrated directly into the actuator and
helps the actuator achieve its desired effect. An example is the angle sensor
in a servo system.
Sometimes the sensor and actuator are the same device. Electric motors,
for example, convert electrical energy into mechanical rotation (actuator), but
they can also be used to convert mechanical rotation into electrical energy for
measurement (sensor). Another example is communication. The sending agent
is performing an action and the receiving agent is sensing. When the receiving
agent responds, the roles reverse.
5.6.2 Image and Signal Processing
To utilize the information supplied by the sensor, some sensor-specific com-
putation is necessary. This computation is used to clean up the information
supplied by the sensor and to apply processing algorithms that detect the
attribute being measured.
Some sensors require little processing. A simple mechanical switch is either
on or off and its interpretation should be easily understood by the agent.
However, mechanical switches bounce when their position is changed, and
this bounce causes a rapid series of on/off signals to be generated before the
switch settles into its new position. The computer systems interpreting signals
from switches are fast enough to detect these bounces and take them into
consideration so as not to misinterpret the environment. Therefore, mechanical
switches are de-bounced using a small electrical circuit or software. This is a
simple example of signal processing.
Other sensors require massive amounts of signal processing. Radar and
sonar systems work in environments with a large amount of background noise.
If the signals are not carefully processed, the background noise will be in-
correctly categorized as an item of interest to the system. Complex, sensor-
specific algorithms are used to eliminate the background noise, interpret the
results, and find target items in the signal.
In a similar manner, imaging detectors require large amounts of postpro-
cessing to clean up device specific noise and interpret the data and find target
items of interest to the agent. In many situations, the algorithms are layered,
where lower level algorithms clean up the individual pixels and higher level
algorithms merge multiple pixels together to interpret what is in the image.
Image and signal processing can be computationally very expensive. Some-
times this computational expense will justify the use of special purpose pro-
cessors whose only purpose is to perform the image or signal processing.
5.6.3 Data Fusion
The ultimate goal of perception is to supply the agent with an accurate repre-
sentation of the working environment. In most real world systems, perception
requires the agent to integrate multiple sensor data into a single consistent
