Online Synthesis Of Speculative Building Information Models For Robot Motion Planning

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Computer and Information Science
Discipline
Subject
Building Information Modeling
Mapping
Robot Perception
SLAM
Computer Sciences
Robotics
Funder
Grant number
License
Copyright date
2021-08-31T20:20:00-07:00
Distributor
Related resources
Author
Shariati, Armon
Contributor
Abstract

Autonomous mobile robots today still lack the necessary understanding of indoor environments for making informed decisions about the state of the world beyond their immediate field of view. As a result, they are forced to make conservative and often inaccurate assumptions about unexplored space, inhibiting the degree of performance being increasingly expected of them in the areas of high-speed navigation and mission planning. In order to address this limitation, this thesis explores the use of Building Information Models (BIMs) for providing the existing ecosystem of local and global planning algorithms with informative compact higher-level representations of indoor environments. Although BIMs have long been used in architecture, engineering, and construction for a number of different purposes, to our knowledge, this is the first instance of them being used in robotics. Given the technical constraints accompanying this domain, including a limited and incomplete set of observations which grows over time, the systems we present are designed such that together they produce BIMs capable of providing explanations of both the explored and unexplored space in an online fashion. The first is a SLAM system that uses the structural regularity of buildings in order to mitigate drift and provide the simplest explanation of architectural features such as floors, walls, and ceilings. The planar model generated is then passed to a secondary system that then reasons about their mutual relationships in order to provide a water-tight model of the observed and inferred freespace. Our experimental results demonstrate this to be an accurate and efficient approach towards this end.

Advisor
Camillo J. Taylor
Date of degree
2020-01-01
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation