Fusion in Robotics - Information Fusion in Signal and Image Processing - page 47

the robot has to be equipped with proprioceptive sensors such as an inertial measure-
ment unit, accelerometers, odometers, etc. In many cases, there are also mechanical
models of the robot available, which make it possible to complete the knowledge of
the system's state. For example, the direct model of a manipulator robot gives us the
position and orientation of its effector (tool) depending on the position value of each
of its axes. Likewise, knowledge of a robot's kinematic model, based on the rotation
speed of the wheels, can be used to evaluate its speed.
Before and during a movement, the robot has to localize itself with respect to its
environment. In order to do this, it is necessary for it to perceive objects using sensors
that are adequate to this environment, which are referred to as exteroceptive sensors.
Under water, for example, the use of sonar will be preferred; for driving at night, an
infrared camera will be used. The topics by [CHA 98, ZHA 97] give a rather thorough
presentation of everything pertaining to perception in intelligent vehicles.
4.2.3. Interaction with the operator and symbolic interpretation
A robot's purpose is to maneuver in space instead of or alongside humans in order
to replace or to help them in achieving their mission. The extent of the interactions
with the human operator vary depending on the cases. For example, when driving an
automobile or teleoperating, the human operator stays in direct control of the robot
(or of the vehicle). The role of the embedded perception system is then to transmit
information to this operator in order to facilitate their control task. On the contrary,
when performing more automated tasks, for example, with welding robots or Martian
robots for which real-time control is impossible, the information perceived is essen-
tially used by the robot itself. The information transmitted to the human operator only
corresponds to status reports on the progress of the mission, on alarms or on the robot's
situation.
Depending on their use (human or robot operator), the semantic nature of the infor-
mation will be quite different, i.e. more symbolic for the human operator (the object
is close, the speed is high), and more numerical if they are directly used by the robot
(distance from the tool point to the closest point on the table = 20 cm).
4.2.4. Time constraints
The robot maneuvers inside the environment at a certain speed. The environment
can also change with time (other robots or vehicles) at varying speeds. In order for the
robot to achieve its mission, the frequency of observation of its state and the state of
its environment needs to be high enough (Shannon's theorem). For safety reasons, it
also has to be able to perceive “events” that require a rapid response or action.