Simultaneous Localization And Mapping SLAM Introduction to the Problem and Approaches

1
Simultaneous Localization And Mapping
(SLAM)Introduction to the Problem and Approaches

• R. Chatila
• LAAS-CNRS
• Toulouse, France
• Raja.Chatila_at_laas.fr

2
The Navigation Problem
How to get to Stand 6?
3
The Navigation problem
Where am I?
How to get to  Twin Peaks ?
4
Navigation

• Need for a map representing the environment
• for localization
• for motion planning

5
The Mapping Problem

• Unknown environment mapping by mobile vehicles
• Incremental.
• Structured-Indoors/Unstructured-Natural.
• Static/Dynamic.
• Noisy sensors
• No external localization means.
• Multisensory perception vision, laser, radar,
• Single/multiple vehicles
• Applications planetary exploration, terrain
  mapping, aerial imagery, ocean floor mapping,

6
Mapping

• Representations
• Appearance, Grids, Features, Objects, Topology.
• Incremental mapping Environment
  exploration, and fusion of local perceptions.
  Explicit processing of uncertainties for building
  a global consistent map.

7
Representations and Maps
Topology
Features
Objects
Appearance
Grids
8
Local Map
Local map
Noisy measurements (2D laser scan)
Piecewise linear approximation uncertainties
9
ConsistentIncremental Mapping

• Mapping requires an accurate position
  reference
• Localization requires an accurate map
• Odometry/IMU has cumulative errors and cannot
  provide a correct position after some travelled
  distance/time

10
Incremental Mapping
Position reference given by odometry map
distortion
Jensfelt
11
Incremental Mapping
?
Jensfelt
12
Incremental Mapping
Closing the loop

• Cumulative position errors.
• External correction by mapped environment
  elements themselves.

Konolige
13
A Classical Problem (1984 - )

• Fox, Thrun, Burgard
• Gutman
• Christensen
• Jensfelt
• Nebot
• Newman
• Leonard
• Durrant-Whyte-ACFR
• Dissanyake
• Rives
• Uhlmann
• Siegwart
• Sukkarieh
• Davison
• Adams
• Lacroix

• Brooks
• Chatila Laumond
• Moravec Elfes
• Crowley
• Faugeras Ayache
• Smith Cheeseman
• Leonard Durrant-Whyte
• Moutarlier Chatila
• Cox
• Devy
• Simmons
• Castellanos
• Tardos
• Neira
• Konolige
• Milios

14
Simultaneous Localization And Mapping An Old
problem
15
Simultaneous Localization And Mapping An Old
problem
16
SLAM Uncertainties
17
SLAM Uncertainty reduction
18
S L A M General Procedure
MOTION
PERCEPTION
LOCAL MAP
SELECT VIEWPOINT
Prediction and Data association problem
GLOBAL MAP
MATCHING
ROBOT POSE CORRECTION
MAP FUSION
19
SLAM Problems and Approaches

• Modeling sensor uncertainties
• Bayesian filtering
• Gaussian PDF Extended Kalman filter.
• Sample-based representations Particle filters.
• The Data Association Problem and multiple
  hypotheses.
• Dealing with complexity hierachical models
  decorrelation.
• Dynamic environments
• Multi-robots, multisensors.

20
A classical problem (1985 - )

• Brooks 85
• Chatila Laumond 85
• Moravec Elfes 85
• Crowley 85, 89, 94
• Faugeras Ayache 86
• Smith Cheeseman 87
• Drumheller 87
• Leonard Durrant-Whyte 88, 91
• Moutarlier Chatila 89,90,91
• Cox 89
• Devy 94
• Simmons 95
• Castellanos 99
• Neira

• Betge 96
• Konolige Millos 98
• Fox, Thrun, Burgard 98-04
• Gutman 00
• Christensen 98-04
• Jensfelt 98,01,02,03
• Nebot 99-03
• Leonard Neuman, 00-03
• Durrant-Whyte 99-04
• Rives 01,02
• Uhlmann 01
• Siegwart 02
• Pradalier 02
• Sukkarieh