Feature-based localization using fixed ultrasonic transducers

Describes an approach for mobile robot localization based on geometric features extracted from ultrasonic data. In previous work the authors (1994) proposed a multi-stage approach to sonar data clustering which extracts planar surface features in real time using transducers fixed relative to the mobile platform. The method is efficient, precise, and robust in the face of both measurement noise and significant dead-reckoning error. In this work the authors apply the feature extraction algorithm to the problem of localization in a partially known environment. The authors describe an approach for establishing correspondences between extracted and map features. A least-squares approach to mobile robot pose estimation is delineated which is linear in the number of extracted features. The decoupling of the correspondence-matching and pose-estimation stages offers advantages in speed and precision. There are no constraints on the trajectory to be followed for localization except that sufficiently large portions of features be ensonified to allow clustering. Preliminary experiments indicate the utility of the approach, especially for accurate orientation estimation.

Authors

• Robert Mandelbaum
• Max Mintz

Keywords

• Ultrasonic transducers
• Feature extraction
• Mobile robots
• Data mining
• Sonar measurements
• Noise robustness
• Noise measurement
• Ultrasonic variables measurement
• Working environment noise
• Clustering algorithms
• Dead Reckoning
• Feature Maps
• Kalman Filter
• Pose Estimation
• Mobile Robot
• Root Mean Square Error Of Cross-validation
• Region Of Attraction
• 2D Case
• Occupancy Grid
• 1D Case
• Planar Features
• Feature Merging