SLAM the Robots

1
SLAM the Robots

• Robotics Research
• Department of Computer Science
• University of Waikato

2
Robot Mapping and Navigation

• Why its so interesting?
• What we are doing?

3
(No Transcript)
4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
Problem 1 LocalisationWhere am I?

• The robot has a map provided along with imperfect
  sensors and imperfect odometry
• Also need to consider control issues - how the
  robot explores and how it navigates

8
Problems environment clutter, people
9
(No Transcript)
10
Problem 2 Localisation and Mapping

• Robot computes its own map from it own experience
  of its environment with its imperfect sensors
  and imperfect odometry
• Robot needs to estimate its location at the same
  time it is estimating its map
• Simultaneous Localisation and Mapping (SLAM)

11
The localisation problem
12
(No Transcript)
13
Localisation

• Sensing and movement errors are a huge problem
• In global representations they accumulate
• Solutions try to
• Minimise the error
• Minimise the effect
• Managing the uncertainty
• Representation is an important issue

14
Closing the Cycle
15
(No Transcript)
16
Why dont we use GPS?
17
Representation

• Global evidence-grid approach

18
Problems- Way too much data Need online (real
time) solutions Cycle Closing Need to propagate
an error correction back through the map
19
Topological Representations
20
Problem Tricky matching problem to detect cycles
21
Our approach

• Combine information from different sources
• Combine topological and absolute mapping
• Topological map is the prime representation
• Detect cycles in a topological map
• Detect cycles in an absolute map
• Augment representation with 3D visual landmarks

22
The Mapping Approach
Feature based approach Apply a line fitting
algorithm to raw laser data Use scan matching
technique to obtain local consistency Problem -
noise
23
The Mapping Approach
Local Space
Topological Map
Global Metric Map
24
Detecting Cycles in a Global Metric Map
25
(No Transcript)
26
(No Transcript)
27
Visual Landmarks
28
(No Transcript)
29
(No Transcript)
30
420 projects

• Implement a localisation algorithm
• Cycle closing in a global map
• Dynamic environments
• Dealing with uncertainty in topological and
  global maps
• Human tracking
• False positives in cycle closing

31
What you need?

• Part 1 Math (preferably Algebra)
• We have some projects of interest to
  Statisticians
• Know how to program
• Enthusiasm