Page 224 - Introduction to AI Robotics

207
6.4 Proprioceptive Sensors
to retrieve the robot and fix it. Sensors are especially important because a
faulty sensor can cause the robot to “hallucinate.” As a result, sensor suites
may offer some sort of redundancy.
PHYSICAL There are two types of redundancy. Physical redundancy means that there
REDUNDANCY are several instances of physically identical sensors on the robot. Fig. 6.4
shows a robot with redundant cameras. In this case, the cameras are mounted
180 apart, and when one sensor fails, the robot has to “drive backwards” in
order to accomplish its task.
LOGICAL Logical redundancy means that another sensor using a different sensing mo-
REDUNDANCY dality can produce the same percept or releaser. For example, the Mars So-
journer mobile robot had a laser striping system for determining the range
to obstacles, which emitted a line of light. If the surface was flat, a cam-
era would see a flat line, whereas if the surface was rough the line would
appear jagged. The robot also carried a second camera which could do ob-
stacle detection via stereo triangulation. The laser striping sensor and the
stereo cameras are logically redundant. They are not physically redundant,
but they produce the same overall information: the location of obstacles rel-
ative to the robot. However, logically redundant sensors are not necessarily
equivalent in processing speed or accuracy and resolution. A stereo range
sensor and algorithm computes the range much slower than a laser striping
system.
Physical redundancy introduces new issues which are the area of active
research investigation. Possibly the most intriguing is how a robot can de-
termine that a sensor (or algorithm) has failed and needs to be swapped out.
FAULT TOLERANCE Surviving a failure is referred to as fault tolerance. Robots can be programmed
in most cases to tolerate faults as long as they can identify when they occur.

6.4 Proprioceptive Sensors

Proprioception is dead reckoning, where the robot measures a signal origi-
nating within itself. In biology, this is usually some measure of how far an
arm or leg has been extended or retracted. In robotics, actuators are generally
SHAFT ENCODER motors. Many motors come with a shaft encoder which measures the number
of turns the motor has made. If the gearing and wheel size is known, then the
number of turns of the motor can be used to compute the number of turns
of the robot’s wheels, and that number can be used to estimate how far the
robot has traveled.