Predicting the performance of ORB-SLAM3 on embedded platforms

Simultaneous Localization and Mapping (SLAM) is a crucial component to the push towards full autonomy of robotic systems, yet it is computationally expensive and can rarely achieve real-time execution speeds on embedded platforms. Therefore, a need exists to  evaluate the performance of SLAM algorit...

Full description

Saved in:
Bibliographic Details
Main Authors: Jacques Matthee, Kenneth Uren, George van Schoor, Corne van Daalen
Format: Article
Language:English
Published: South African Institute of Computer Scientists and Information Technologists 2024-12-01
Series:South African Computer Journal
Subjects:
Online Access:https://sacj.org.za/article/view/20099
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850127174996066304
author Jacques Matthee
Kenneth Uren
George van Schoor
Corne van Daalen
author_facet Jacques Matthee
Kenneth Uren
George van Schoor
Corne van Daalen
author_sort Jacques Matthee
collection DOAJ
description Simultaneous Localization and Mapping (SLAM) is a crucial component to the push towards full autonomy of robotic systems, yet it is computationally expensive and can rarely achieve real-time execution speeds on embedded platforms. Therefore, a need exists to  evaluate the performance of SLAM algorithms in practical embedded environments – this paper addresses this need by creating  prediction models to estimate the performance that ORB-SLAM3 can achieve on embedded platforms. The paper uses three embedded platforms: Nvidia Jetson TX2, Raspberry Pi 3B+ and the Raspberry Pi 4B, to generate a dataset that is used in training and  testing performance prediction models. The process of profiling ORB-SLAM3 aids in the selection of inputs to the prediction model as  well as benchmarking the embedded platforms’ performances by using PassMark. The EuRoC micro aerial vehicle (MAV) dataset is used to generate the average tracking time that the embedded platforms can achieve when executing ORB-SLAM3, which is the target  of the prediction model. The best-performing model has the following results 2.84%, 3.93%, and 0.95 for MAE, RMSE and R2 score  respectively. The results show the feasibility of predicting the performance that SLAM applications can achieve on embedded  platforms.
format Article
id doaj-art-00ed506f8f294a99ad769ffa36eeb35e
institution OA Journals
issn 1015-7999
2313-7835
language English
publishDate 2024-12-01
publisher South African Institute of Computer Scientists and Information Technologists
record_format Article
series South African Computer Journal
spelling doaj-art-00ed506f8f294a99ad769ffa36eeb35e2025-08-20T02:33:44ZengSouth African Institute of Computer Scientists and Information TechnologistsSouth African Computer Journal1015-79992313-78352024-12-0136210.18489/sacj.v36i2/2009924517Predicting the performance of ORB-SLAM3 on embedded platformsJacques Matthee0https://orcid.org/0000-0002-2139-0648Kenneth Uren1https://orcid.org/0000-0002-0561-0735George van Schoor2https://orcid.org/0000-0001-5702-1812Corne van Daalen3https://orcid.org/0000-0002-9849-586XNorth-West UniversityNorth-West UniversityNorth-West UniversityStellenbosch UniversitySimultaneous Localization and Mapping (SLAM) is a crucial component to the push towards full autonomy of robotic systems, yet it is computationally expensive and can rarely achieve real-time execution speeds on embedded platforms. Therefore, a need exists to  evaluate the performance of SLAM algorithms in practical embedded environments – this paper addresses this need by creating  prediction models to estimate the performance that ORB-SLAM3 can achieve on embedded platforms. The paper uses three embedded platforms: Nvidia Jetson TX2, Raspberry Pi 3B+ and the Raspberry Pi 4B, to generate a dataset that is used in training and  testing performance prediction models. The process of profiling ORB-SLAM3 aids in the selection of inputs to the prediction model as  well as benchmarking the embedded platforms’ performances by using PassMark. The EuRoC micro aerial vehicle (MAV) dataset is used to generate the average tracking time that the embedded platforms can achieve when executing ORB-SLAM3, which is the target  of the prediction model. The best-performing model has the following results 2.84%, 3.93%, and 0.95 for MAE, RMSE and R2 score  respectively. The results show the feasibility of predicting the performance that SLAM applications can achieve on embedded  platforms.https://sacj.org.za/article/view/20099monocular-inertial slamorb-slam3embedded platformnvidia jetson tx2raspberry pi
spellingShingle Jacques Matthee
Kenneth Uren
George van Schoor
Corne van Daalen
Predicting the performance of ORB-SLAM3 on embedded platforms
South African Computer Journal
monocular-inertial slam
orb-slam3
embedded platform
nvidia jetson tx2
raspberry pi
title Predicting the performance of ORB-SLAM3 on embedded platforms
title_full Predicting the performance of ORB-SLAM3 on embedded platforms
title_fullStr Predicting the performance of ORB-SLAM3 on embedded platforms
title_full_unstemmed Predicting the performance of ORB-SLAM3 on embedded platforms
title_short Predicting the performance of ORB-SLAM3 on embedded platforms
title_sort predicting the performance of orb slam3 on embedded platforms
topic monocular-inertial slam
orb-slam3
embedded platform
nvidia jetson tx2
raspberry pi
url https://sacj.org.za/article/view/20099
work_keys_str_mv AT jacquesmatthee predictingtheperformanceoforbslam3onembeddedplatforms
AT kennethuren predictingtheperformanceoforbslam3onembeddedplatforms
AT georgevanschoor predictingtheperformanceoforbslam3onembeddedplatforms
AT cornevandaalen predictingtheperformanceoforbslam3onembeddedplatforms