Fusion in Robotics - Information Fusion in Signal and Image Processing - page 50

We know the map's localization with respect to the universe frame T R map
R u

as well as

the localization of the walls with respect to the map T R wall i
R map

. The sensors give us the

localization of the walls with respect to the sensors T R wall i
R sensor j

and the calibration pro-
vides us with the localization of the sensors with respect to the frame associated with
the robot T R sensor j
R ( k )

. We can then infer that:

= T R map
R u

T R wall i
R map

T R wall i
R sensor j

− 1 T R sensor j

R ( k )

− 1

T R ( k )
R u

· · ·

map

R

wall 2

R

wall 1

R

sensor2

R

sensor1

R(k)

Ru R

map

Figure 4.1. Localization of a moving robot in its environment which is described in a map

This simple example underlines the typical problems that have to be solved in
robotics:
- the determination of transformations T R sensor j
R ( k )

by calibration;

- the evaluation of the transformation T R wall i
R sensor j

based on sensor information,

which requires signal and image processing;
- matching the “walls” that are seen with those that are known and on the map;
- the fusion of information from different sensors that see the various walls.

4.3.2. Types and levels of representation of the environment

In order to define a strategy for action or perception, it is often necessary to rely on
an interpretation of the situation based on symbolic data with a high semantic level.
For example, we can be faced with situations such that:
- the effector is close to the pin;