Curiosity Driven Exploration with Focused Semantic Mapping
Date
2019
Authors
Contributor
Advisor
Department
Instructor
Depositor
Speaker
Researcher
Consultant
Interviewer
Narrator
Transcriber
Annotator
Journal Title
Journal ISSN
Volume Title
Publisher
University of Hawaii at Manoa
Volume
Number/Issue
Starting Page
Ending Page
Alternative Title
Abstract
Among many similar natural creatures, humans have the impressive ability to seamlessly explore a new environment and build up their knowledge base. By combining complex exteroceptive perceptions we can “easily” identify the terrain, objects, or hazards around us allowing us to determine a target of attention to make behavioral decisions accordingly. However, sophisticated exploration behaviors are still nontrivial to today’s mobile robots, despite the recent progress in sensing and processing capabilities. Generally, robots face many challenges in achieving human-level exploration, including sufficient perception for navigation, intelligent knowledge extraction and representation, and autonomous decision- making that maximizes both perception and learning. Semantic description is a promising technique to assist mobile robots to sparsely represent and accumulate knowledge of their environments. In this project, we propose a hybrid focus model scheme using curiosity- driven exploration coupled with a prior knowledge base. In this scheme, the robot uses realtime topic modeling to discriminate the sensing data, with a one-stage feature detector to identify the most curious feature in the scene. Using this topic modeling based curiosity, focus driven adaptive sampling and navigation are implemented to facilitate more robust decisions and maximize the perception within the focused region. The proposed hybrid focus combined with metric information allows the creation of semantic maps that enhance human and robot interaction. Using both underwater video datasets and data collected by a mobile robot platform with a RGB camera in a laboratory setting, the performance of the proposed approach is demonstrated and analyzed.
Description
Keywords
Robotics
Citation
Extent
Format
Geographic Location
Time Period
Related To
Related To (URI)
Table of Contents
Rights
Rights Holder
Local Contexts
Collections
Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.