Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps
Mulubika, C. 2025. Global Positioning System Denied Navigation of Visual Simultaneous Localisation and Mapping Generated Maps. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/d2c63575-e099-4ffa-a42e-5c0e1bf24854
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614079173001216 |
|---|---|
| access_status_str | Open Access |
| author | Mulubika, Chibaye |
| author2 | Schreve, Kristiaan |
| author_browse | Mulubika, Chibaye Schreve, Kristiaan |
| author_facet | Schreve, Kristiaan Mulubika, Chibaye |
| author_sort | Mulubika, Chibaye |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Mulubika, C. 2025. Global Positioning System Denied Navigation of Visual Simultaneous Localisation and Mapping Generated Maps. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/d2c63575-e099-4ffa-a42e-5c0e1bf24854 |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132261 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:46:20.037Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/132261 Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps Mulubika, Chibaye Schreve, Kristiaan Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering. Global Positioning System Mobile robots -- Automatic control Blind automobile mechanics Navigation -- Data processing UCTD Mulubika, C. 2025. Global Positioning System Denied Navigation of Visual Simultaneous Localisation and Mapping Generated Maps. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/d2c63575-e099-4ffa-a42e-5c0e1bf24854 Thesis (PhD)--Stellenbosch University, 2025. ENGLISH ABSTRACT: This dissertation is centred on how a mobile robot platform can make use of visual simultaneous localization and mapping (vSLAM) to get back to a required position in case of camera malfunction or loss of visual capabilities during its mission. It assumes that the environment in which it is operating is both global positioning system (GPS) denied and static. With loss of visual capabilities or camera malfunction in vSLAM, the robot has limited sources of sensor data. Here it is assumed that only the inertial measurement unit (IMU) sensors remain when visual input is not present. Hence data previously generated during vSLAM activities when visual input was available provide an attractive solution to get anywhere in the previously traversed path. To select data needed for blind navigation, the path is treated as a signal that needs to be reconstructed were ego-motion from IMU sensors and image data are the primary source of information. Image data guides the ego-motion along safe passages while ego-motion parameters are periodically sampled during vSLAM map generation. From ego-motion, twist velocities, distance covered, the sampling period and yaw angles are used as path landmarks to describe the path. The yaw angles are further used in segregating path segments into straight and curved segments by assigning them two different sampling periods, one half of the other, with the straight getting the higher value. The path landmarks are stored in a YAML file and uniquely labelled from the start to the point of camera malfunction or loss of visual capabilities. The sampling, storage, and use of the path landmarks use the robot operating system (ROS) software architecture. The publisher-subscriber and action server and client are used to save and use stored data respectively. During back travel, when there is no visual input, the robot uses the saved path landmarks starting with the last saved to get to the base station using two different motion control methods (sampled time-based and travelled distance-based motion control method) Assuming suitable sensors available in addition to vSLAM, the layer can be used as a final resort to navigate the robot to a recovery position when visual input ceases during vSLAM activities. The solution was evaluated on a Turtlebot2i equipped with an RGB-D camera, ROS kinetic software, and RTAB-Map vSLAM on ubuntu 16.04 LTS. Results show that a robot can use selected vSLAM generated data for blind navigation with minimal inconvenience to vSLAM operations. Furthermore, position errors and drift per meter become less significant as the speed for operating increased using sampled time-based motion control while travelled distance-based motion control registered significantly higher position error and drift. Given the remaining sensors used for navigation, which are prone to errors, the results indicate the solution can be applied as a last resort to autonomously navigate the blind robot to its base station over short distances. The solution can be applied to robots where human assistance for retrieval is unavailable or undesirable. The solution is further proposed as a layer to existing ROS implemented vSLAM systems. AFRIKAANSE OPSOMMING: Hierdie proefskrif stel ʼn metode voor vir mobiele robotte om terug te keer na ʼn basis posisie deur van visuele gelyktydige lokalisering en kartering (vSLAM) gegenereerde kaartdata gebruik te maak nadat die kamera gefaal het of daar ʼn verlies van visuele vermoëns was. Dit neem aan dat die omgewing waarin dit werk, beide sonder ʼn globale posisioneringstelsel (GPS) en staties is. Met verlies aan visuele vermoëns of kamera faling in vSLAM, het die robot beperkte bronne van sensordata. Hier word aanvaar dat slegs die traagheidsmetingseenheid (IMU) sensors beskikbaar is wanneer visuele insette nie teenwoordig is nie. Gevolglik bied data wat voorheen tydens vSLAMaktiwiteite gegenereer is toe visuele insette beskikbaar was 'n aantreklike oplossing om by enige plek in die voorheen deurkruis pad te kom. Om data te kies wat nodig is vir blinde navigasie, word die pad hanteer as 'n sein wat gerekonstrueer moet word waar self-beweging en beelddata die primêre bron van inligting is. Beelddata lei slegs die self-beweging in veilige gedeeltes terwyl self-beweging-parameters periodiek gemonster word tydens vSLAMkaartgenerering. Van self-beweging word kronkelsnelhede, afstand afgelê, die dataversamlingsperiode en swaaihoeke as landmerke gebruik om die pad te beskryf. Die swaaihoeke word verder gebruik om padsegmente in reguitlyn en geboë segmente te skei deur aan hulle twee verskillende dataversamelingsperiodes toe te ken, waarvan die waarde van een die helfte van die ander is, met die reguitlyn wat die hoogste waarde kry. Die pad-landmerke word in 'n YAML-lêer gestoor en uniek gemerk van die begin tot die punt van kamera faling of verlies aan visuele vermoëns. Die monsterneming, berging en gebruik van die landmerke gebruik die robotbedryfstelsel (ROS) sagteware-argitektuur. Die uitgewer-intekenaar en aksiebediener en kliënt word gebruik om onderskeidelik gestoorde data te stoor en te gebruik. Tydens die herstel beweging, wanneer daar geen visuele insette is nie, gebruik die robot die gestoorde landmerke wat begin met die laaste gestoorde om by die basisstasie uit te kom deur twee verskillende bewegingsbeheermetodes te gebruik (dataversamelingperiode-gebaseerde en gereisde afstand-gebaseerde bewegingsbeheermetodes). Met die aanname dat geskikte sensore addisioneel tot vSLAM beskikbaar is, kan die laag as 'n finale uitweg gebruik word om die robot te navigeer na ‘n herstelposisie wanneer visuele insette tydens vSLAM-aktiwiteite ophou. Die oplossing is getoets op 'n Turtlebot2i toegerus met 'n RGB-D kamera, ROS kinetiese sagteware, en RTAB-Map vSLAM op ubuntu 16.04 LTS. Resultate toon dat 'n robot geselekteerde vSLAM-gegenereerde data kan gebruik vir blinde navigasie met die minste ongerief vir vSLAM-bedrywighede. Verder word posisiefoute en dryf minder betekenisvol aangesien die spoed vir bedryf toegeneem het deur gebruik te maak van gemonsterde tydgebaseerde bewegingsbeheer terwyl gereisde afstand-gebaseerde bewegingsbeheer beduidend hoër posisiefout en dryf geregistreer het. Gegewe die oorblywende sensors wat vir navigasie gebruik word, wat geneig is tot foute, dui die resultate aan dat die oplossing as 'n laaste uitweg toegepas kan word om die blinde robot op kort afstande outonoom na sy basisstasie te navigeer. Die oplossing kan toegepas word op robotte wat geen menslike hulp nodig het vir herwinning waar daar 'n kritieke tekort aan personeel in binnenshuise omgewings is nie. Die oplossing word verder voorgestel as 'n laag vir bestaande ROS-geïmplementeerde vSLAM-stelsels. Doctoral 2025-06-02T07:13:29Z 2025-06-02T07:13:29Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132261 en Stellenbosch University xviii, 181 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Global Positioning System Mobile robots -- Automatic control Blind automobile mechanics Navigation -- Data processing UCTD Mulubika, Chibaye Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title | Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title_full | Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title_fullStr | Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title_full_unstemmed | Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title_short | Global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| title_sort | global positioning system denied navigation of visual simultaneous localisation and mapping generated maps |
| topic | Global Positioning System Mobile robots -- Automatic control Blind automobile mechanics Navigation -- Data processing UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132261 |
| work_keys_str_mv | AT mulubikachibaye globalpositioningsystemdeniednavigationofvisualsimultaneouslocalisationandmappinggeneratedmaps |