The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #94144

Title:Position Refinement for a Navigating Robot Using Motion Information Based on Honeybee Localisation Strategies
Date: 1994
Abstract:This thesis describes an algorithm for the position refinement of a robot using visual information. The algorithm was inspired by techniques used by navigating honey bees. The idea is to investigate how the bee could exploit the optical flow on its retina induced by its own motion for the global localisation process. The algorithm is based on separating close and distant landmarks in a scene using their different optical flow. We also use the optical flow to construct a map of the close landmarks, which represents their two-dimensional layout in the scene. The algorithm remembers a representation of both close and distant landmarks at a location it later wants to return to. On return to this position, it compares the information gained at the current location with the remembered images. Using the images of the distant landmarks, it is possible to keep the orientation, under which the scene is observed, constant. Then it is possible to calculate a vector pointing to the target position using the nearby landmarks. The algorithm was implemented and tested on a RTX-robot.

[Search These Pages] [DAI Home Page] [Comment]