Localisation

Parent: [SLAM Index](SLAM Index.md)

Source: [Wikipedia Lokalisierung](Wikipedia Lokalisierung.md) The positioning of an autonomous mobile robot relative to its environment

• The position of a mobile robot is seldom known exactly
• An unknown initial position / measurement uncertainties while moving
• Becomes a SLAM problem when neither the position nor the map is known

Goal/Output: POSE

• Due to uncertainties etc, it’s good to have a POSE representation that also shows these uncertainties
• e.g. probability densities, particle clouds

Approaches mostly fusion-based (odometry/sensors + landmarks)

• Cross-bearing known landmarks
• Template-matching current sensor measurements (auch Scan-Matching)
• Probabilistic methods

Local and global localisation

• Local
  • current POSE in the environment is known
  • correct the incremental odometry error that occurs every step
• Global
  • current POSE in the environment is unknown
  • position errors aren’t negligible
  • error of the initial estimated position can be arbitrarily huge
  • the robot has to determine its position first through finding significant landmarks, only then can a local localisation be carried out
• Kidnapped robot problem (check robustness)