Position Estimaton Using Equidistance Lines

An approach to perception-based position estimation is presented. The approach employs a representation of positional information which is based on a global, curvilinear coordinate system which is grounded in the real world. This coordinate system is spanned by the contours of visible objects and by so-called equidistance lines which are constructed from the object contours. Equidistance lines can be computed from arbitrarily curved contours and do not require an analytical representation of these contours. The main advantage of a spatial representation based on equidistance lines is that it does not depend on the existence of salient spatial features in the environment. Furthermore, the global curvilinear coordinate system allows a straightforward transition between the local frames of reference associated with local perceptions. To estimate the position of a robot vehicle over a sequence of local sensor images requires only a minimal number of object contours to be visible or partially visible in these images and a qualitative correspondence of these object contours to be maintained.

Authors

• Erwin Prassler
• Evangelos E. Milios

Keywords

• Vehicles
• Robot kinematics
• Image sensors
• Mobile robots
• Robot sensing systems
• Data mining
• Feature extraction
• Robotics and automation
• Humans
• Animals
• Coordinate System
• Kalman Filter
• Autonomous Vehicles
• Reference System
• Local Image
• Algebraic Equations
• Visual Object
• Position Estimation
• Vehicle Position
• Local Perceptions
• Global Coordinate System
• Curvilinear Coordinates
• Automated Guided Vehicles
• Object Contour
• Local Reference Frame
• Kind Of Perception
• Straight Line
• Minimum Distance
• Local System
• Image Segmentation
• Distance Map
• Local Coordinate System
• Local Coordinate
• Binary Image
• Reference Image
• Distance Vector
• Local Setting
• Distance Values
• Current Position
• Edge Pixels