Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads
Thesis (MEng)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA en_ZA |
| Published: |
Stellenbosch : Stellenbosch University,
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613929300033536 |
|---|---|
| access_status_str | Open Access |
| author | Hohbach, Annika |
| author2 | Jordaan, Willem |
| author_browse | Hohbach, Annika Jordaan, Willem |
| author_facet | Jordaan, Willem Hohbach, Annika |
| author_sort | Hohbach, Annika |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/127072 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA en_ZA |
| last_indexed | 2026-06-10T12:43:57.159Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University, |
| publisherStr | Stellenbosch : Stellenbosch University, |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/127072 Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads Hohbach, Annika Jordaan, Willem Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. Payloads (Aerospace engineering) Drone aircraft Aerodynamic load Thesis (MEng)--Stellenbosch University, 2023. ENGLISH ABSTRACT: This thesis addresses the viability of a model-free trajectory planning solution in comparison to popular model-based solutions found in the literature for a rotary-wing unmanned aerial vehicle with variable suspended payloads. The varied parameters consist of different payload masses and cable lengths. The oscillating suspended payload introduces unpredictable dynamics to the system that are challenging to control or plan for. Through research, four suitable trajectory planning solutions are selected to be implemented as the local planner and verified in simulation. A baseline is defined that only consists of the onboard proportional-integral-derivative controllers to track global waypoints. The baseline shows large swing angles and instabilities, as well as a lack of robustness to payload uncertainties. The four chosen methods add additional trajectory planning to the global path in order to stabilise flight. All methods are compared to the baseline and each other with a non-linear mathematical model as well as a Software-in-the-loop Gazebo simulation on three defined circuits that aim to expose different characteristics and limitations of each method. Robustness tests are also conducted by varying the payload parameters. The first method uses a notch filter as an input shaper to generate swing-reducing trajectories. It is the simplest method to implement and reduces the swing angles from the baseline. It does have limitations because it does not consider the entire path. The next method uses an offline-generated cubic spline and a pure pursuit tracker to provide better and more flexible tracking. It is the least computationally expensive method and is seen to produce desirable results on simple circuits, but its performance degrades on complex circuits. A non-linear model predictive controller (NMPC) is applied as a more adaptable solution that allows for constraints. Iteratively solving at each time step with unpredictable solving times makes it a computationally expensive method, but it has been shown to be the best model-based solution since it can be tailored to the different circuits and produces good results for each. A twin-delayed deep deterministic policy gradient (TD3) reinforcement learning (RL) agent is trained as the model-free solution. It is less computationally expensive than the NMPC as it does not need to solve iteratively and allows for input commands to be executed deterministically. The TD3 agent delivers comparable results and the robustness and performance improves when trained with changing payload parameters. A reduced model is used for accelerating the training process with the trained agent being transferable to two different testing environments. The use of RL-based model-free solutions for trajectory planning is suitable to stabilise the flight of a RUAV with a suspended payload and works well in the face of uncertainty. AFRIKAANS OPSOMMING: Hierdie tesis spreek die lewensvatbaarheid van ’n modelvrye trajekbeplanningsoplossing aan in vergelyking met gewilde modelgebaseerde oplossings in die literatuur vir ’n roterende-vlerk onbemande lugvoertuig met veranderende swaaiende loonvragte. Die veranderende parameters bestaan uit die loonvragmassas en kabellengtes. Die ossillerende loonvrag lei tot onvoorspelbare dinamika in die stelsel wat uitdagend is om te beheer of voor te beplan. Vier geskikte rajekbeplanningsoplossings is gekies om as die plaaslike beplanner ge¨ımplementeer te word en in simulasie geverifieer te word. ’n Basislyn word gedefinieer en bestaan slegs uit die proporsionele-integraal-afgeleide beheerders aan boord om die globale wegpunte op te volg. Die basislyn toon groot swaaihoeke en onstabiliteite, sowel as ’n gebrek aan robuustheid, wanneer verskillende loonvragte getoets word. Die vier gekose metodes voeg bykomende trajekbeplanning by die globale pad om die vlug te stabiliseer. Alle metodes word met die basislyn en mekaar vergelyk met ’n nie-lineˆere wiskundige model sowel as ’n Sagteware-in-die-lus Gazebo simulasie op drie gedefinieerde roetes, wat daarop gemik is om verskillende eienskappe en beperkings van elke metode bloot te stel. Robuustheidstoetse word ook uitgevoer deur die loonvragparameters te verander. Die eerste metode gebruik ’n bandbeperkfilter as ’n intreefilter om swaaiverminderende trajekte te genereer. Dit is die eenvoudigste metode om te implementeer en verminder die swaaihoeke vanaf die basislyn. Dit het beperkings omdat dit nie die hele pad in ag neem nie. Die volgende metode gebruik ’n vanlyn-gegenereerde kubieke kurwe en ’n suiwer strewe-spoorsnyer om beter en meer buigsame padvoltooiing te bied. Dit is die kortse berekeningsduurmetode en word gesien om gewenste resultate op eenvoudige roetes te lewer, maar sy werkverrigting verswak op komplekse roetes. ’n Nie-lineˆere modelvoorspellendekontroleerder (NMPC) word toegepas as ’n meer aanpasbare oplossing wat voorsiening maak vir beperkings. Om iteratief op te los by elke tydstap met onvoorspelbare oplostye maak dit ’n berekeningsduurmetode, maar dit is getoon dat dit die beste modelgebaseerde oplossing is aangesien dit aangepas kan word vir die verskillende roetes en goeie resultate vir elkeen lewer. ’n weelingvertraagde diep deterministiese beleidsgradi¨ent (TD3) versterkingsleeragent word opgelei as die modelvrye oplossing. Dit is minder berekeningsduur as die NMPC aangesien dit nie iteratief hoef op te los nie en dit moontlik maak dat invoeropdragte deterministies uitgevoer kan word. Die TD3-agent lewer vergelykbare resultate en die robuustheid en werkverrigting verbeter met opgeleiding wat gebruik maak van veranderende loonvragparameters. ’n Vereenvoudigde model word vir versnelling van die opleidingsproses gebruik, met die opgeleide agent wat oordraagbaar is na twee verskillende toetsomgewings. Die gebruik van RL-gebaseerde modelvrye oplossings vir trajekbeplanning is geskik om die vlug van ’n RUAV met ’n swaaiende loonvrag te stabiliseer en werk goed in die lig van onsekerheid. Masters 2023-03-03T07:48:16Z 2023-05-18T07:02:54Z 2023-03-03T07:48:16Z 2023-05-18T07:02:54Z 2023-03 Thesis http://hdl.handle.net/10019.1/127072 en_ZA en_ZA Stellenbosch University xvi, 109 pages : illustrations application/pdf Stellenbosch : Stellenbosch University, |
| spellingShingle | Payloads (Aerospace engineering) Drone aircraft Aerodynamic load Hohbach, Annika Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title | Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title_full | Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title_fullStr | Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title_full_unstemmed | Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title_short | Trajectory planning methods applied to a rotary wing UAV with uncertain suspended payloads |
| title_sort | trajectory planning methods applied to a rotary wing uav with uncertain suspended payloads |
| topic | Payloads (Aerospace engineering) Drone aircraft Aerodynamic load |
| url | http://hdl.handle.net/10019.1/127072 |
| work_keys_str_mv | AT hohbachannika trajectoryplanningmethodsappliedtoarotarywinguavwithuncertainsuspendedpayloads |