Introduction to
Probabilistic
Robot Mapping
What is Robot Mapping?
General Definitions for robot
mapping
Terms and concepts related to
Robot Mapping
What is SLAM?
Example of Localization for a mobile robot
Yellow means fixed
firm information
Predicted state
• Robot knows map
• Robot knows landmarks on map
• Robot sees landmarks
• Robot wants to estimate its pose
Example of Mapping Robot creates
the map
estimate
given
• Robot does not know the map or its part
• Robot knows its pose
• Robot sees landmarks
• Robot wants to estimate landmarks on the map to create or
update or extend the map.
Example of SLAM
Predicted value
Real value
• Robot does not know the map or its part
• Robot estimates its pose
• Robot sees landmarks
• Robot wants to estimate landmarks on the map to create or
update or extend the map.
The SLAM problem is chicken-or-egg problem
SLAM Problem is very important
• SLAM is the
fundamental problem
in robot navigation.
• You cannot avoid it.
Applications of SLAM
• In MCECSBOT we do not have SLAM as the map is known. SLAM can be used for
furniture only and items that are not on a map of the building
Formal
Definition of the
SLAM Problem
Definition of the SLAM Problem
All our work is based on
Probabilistic Approaches
Representation of robot’s
uncertainty in probabilistic terms
• We use the same notation as in past lectures
Graphical Model of Full SLAM
path
controls
observations
map
Full SLAM versus Online
SLAM
FULL SLAM
Let us compare
full SLAM and
Online SLAM
Graphical Model
of Online SLAM
Online SLAM
Graphical Model of Online SLAM
to explain the integrations
Why SLAM problem is so
hard to solve?
• The problem can be solved
because map and pose
estimates are correlated
Why SLAM is a hard problem to solve?
• More reasons why it is so hard.
Taxonomy of
SLAM
problems
In active SLAM we have a
feedback to make decision
where to go next
Time is restricted Space is
restricted
Approaches to solve the
SLAM problem
Main Paradigms
for SLAM
Models for
SLAM
Model of Motion and Observation
Model of Motion for SLAM
Examples of Models of Motion
STANDARD ODOMETRY Model for motion of a robot
• Calculate
new data
from old
data and
controls
Model of Observation of Sensor
Examples of Observation Model
Summary on SLAM