Nothing can happen in any
system with any
accuracy unless there is some feedback, and a
sensor of some kind is used to provide that feedback.
In signal applications, devices such as those referenced by Wiccanpiper's excellent writeup in the detector node are also used as sensors to provide a reference to the control circuit of whatever system involved. The feedback is used as a reference to the original signal as a check to see how accurately the signal was carried through the system.
Feedback is vital to any control system. For example, in a power supply that drives a circuit, in order to provide proper current regulation there needs to be a feedback reference at the point of load to ensure that the current fed into the circuit maintians a sufficient level (just enough, not tooo much or not enough), feedback is provided backwards as a reference to tell the control circuit that the amount of current fed into the front end creates the proper amount of current at the back end.
In a robotic motion application, sensors provide the feedback that the nerves in a persons hand and finger provide to ensure that the pressure given is just the right amount. That is how a sophisticated robotic arm can pick up an egg. To simply set the robot's "hand" for a certain pressure isn't enough, there has to be a pressure feedback to compensate for differences in the thickness of the shell.
In an audio circuit, the feedback makes sure that the signal isn't over- or under-driven, and in the case of some subwoofers, there may be an actual physical motion sensor attached to the speaker cone itself to measure its movement (Velodyne makes a great example) to ensure that the signal that started makes it to the end exactly as intended (there is a school of audiophile thought that feels that excessive feedback makes the sound too sterile.)
However you slice it, unless you really, really know what is coming out of the other end you can never be in control of whatever system you create.