Edge-server visual localization for autonomous agents

Simultaneous Localization and Mapping (SLAM) of an unfamiliar environment is an important component of mobile autonomous robots and navigation systems. The autonomous agent must be able to locate themselves effectively in both indoor and outdoor situations. However, indoor SLAM cannot rely on Global...

Full description

Saved in:
Bibliographic Details
Main Author: Tay, Kee Kong
Other Authors: Lam Siew Kei
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2021
Subjects:
Online Access:https://hdl.handle.net/10356/153291
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-153291
record_format dspace
spelling sg-ntu-dr.10356-1532912021-11-16T08:39:16Z Edge-server visual localization for autonomous agents Tay, Kee Kong Lam Siew Kei School of Computer Science and Engineering ASSKLam@ntu.edu.sg Engineering::Computer science and engineering::Computer systems organization::Special-purpose and application-based systems Simultaneous Localization and Mapping (SLAM) of an unfamiliar environment is an important component of mobile autonomous robots and navigation systems. The autonomous agent must be able to locate themselves effectively in both indoor and outdoor situations. However, indoor SLAM cannot rely on Global Positional System (GPS), and existing methods frequently assume static scene circumstances that are unlikely to occur in real-world scenarios. Furthermore, the maps created by traditional SLAM systems have no semantic significance of the environment, where the latter is useful for higher-level missions like path navigation or object classification/detection. The computation cost is another important consideration as SLAM systems are often implemented on embedded platforms. The goal of this project is to develop an edge-server collaborative visual-inertial semantic localization system. A computer vision-based positioning system (visual-inertial SLAM system) is designed on the Jetson Xavier NX embedded platform that can work in dynamic environments and generate context-aware sparse 3D semantic segmentation. The SLAM system on the Jetson Xavier NX communicates the computed locations to a server, which will produce a global map. The SLAM system also obtains inertial information from an Inertial Measurement Unit (IMU) sensor to reduce uncertainty in pose estimation, hence enabling robust pose estimation. Finally, the SLAM system relies on a deep learning module to obtain semantic information for building global semantic maps. Bachelor of Engineering (Computer Engineering) 2021-11-16T04:10:44Z 2021-11-16T04:10:44Z 2021 Final Year Project (FYP) Tay, K. K. (2021). Edge-server visual localization for autonomous agents. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/153291 https://hdl.handle.net/10356/153291 en SCSE20-0742 application/pdf Nanyang Technological University
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Engineering::Computer science and engineering::Computer systems organization::Special-purpose and application-based systems
spellingShingle Engineering::Computer science and engineering::Computer systems organization::Special-purpose and application-based systems
Tay, Kee Kong
Edge-server visual localization for autonomous agents
description Simultaneous Localization and Mapping (SLAM) of an unfamiliar environment is an important component of mobile autonomous robots and navigation systems. The autonomous agent must be able to locate themselves effectively in both indoor and outdoor situations. However, indoor SLAM cannot rely on Global Positional System (GPS), and existing methods frequently assume static scene circumstances that are unlikely to occur in real-world scenarios. Furthermore, the maps created by traditional SLAM systems have no semantic significance of the environment, where the latter is useful for higher-level missions like path navigation or object classification/detection. The computation cost is another important consideration as SLAM systems are often implemented on embedded platforms. The goal of this project is to develop an edge-server collaborative visual-inertial semantic localization system. A computer vision-based positioning system (visual-inertial SLAM system) is designed on the Jetson Xavier NX embedded platform that can work in dynamic environments and generate context-aware sparse 3D semantic segmentation. The SLAM system on the Jetson Xavier NX communicates the computed locations to a server, which will produce a global map. The SLAM system also obtains inertial information from an Inertial Measurement Unit (IMU) sensor to reduce uncertainty in pose estimation, hence enabling robust pose estimation. Finally, the SLAM system relies on a deep learning module to obtain semantic information for building global semantic maps.
author2 Lam Siew Kei
author_facet Lam Siew Kei
Tay, Kee Kong
format Final Year Project
author Tay, Kee Kong
author_sort Tay, Kee Kong
title Edge-server visual localization for autonomous agents
title_short Edge-server visual localization for autonomous agents
title_full Edge-server visual localization for autonomous agents
title_fullStr Edge-server visual localization for autonomous agents
title_full_unstemmed Edge-server visual localization for autonomous agents
title_sort edge-server visual localization for autonomous agents
publisher Nanyang Technological University
publishDate 2021
url https://hdl.handle.net/10356/153291
_version_ 1718368073314992128