Date of Award
1993
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Abstract
The Occupancy Grid Method is a probabilistic spatial modelling technique. In this method, the space to be navigated is subdivided into a grid of cells. Associated with each cell is a probability that indicates the likelihood that that cell is occupied. This method was developed for handling a world where the robot was the only moving object. Thus, it lacks the ability to predict the location of moving objects over time.;In my thesis, an Extended Occupancy Grid Method is presented. The extension handles objects moving at constant velocity. In this method, a four-dimensional location-velocity space is subdivided into cells. Each cell (location-velocity combination) is associated with a probability indicating the likelihood that there is an object at that location moving at that velocity. This probabilistic information is updated over time incrementally using the Bayesian reasoning formula. To calculate such a probability, two pieces of probabilistic information are considered, one indicating the likelihood that there are objects at that location and the other indicating the likelihood that an object at a given location moves with a particular velocity. Methods are presented for deriving the probability for each cell in the extended occupancy grid being occupied by combining the original Occupancy Grid Method with motion detection mechanisms such as the Hough transform.;The Hough transform, originally formulated for recognizing lines in a two dimensional space, has been used to recognize lines in a three-dimensional space for identifying initial location and motion information of objects from sensor data. The approach of combining the Occupancy Grid Method with the Hough transform is robust in the case of occlusion. It can be easily extended to handle other motions, such as rotation and acceleration.
Recommended Citation
Zhang, Yuefeng, "Mobile Robot Perception In Unknown And Unstructured Dynamic Voxel Environments" (1993). Digitized Theses. 2276.
https://ir.lib.uwo.ca/digitizedtheses/2276