Motion planning for a rotary-wing UAV in dynamic environments

Date
2020-03
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: In order for a fully autonomous unmanned aerial vehicle (UAV) to navigate safely in dynamic environments, a conflict free trajectory between an initial and goal state of a vehicle needs to be planned quickly and effectively. Most trajectory planning methods today only consider static environments and do not act on environmental changes that might occur. In order for a UAV to navigate through dynamic environments, a motion planning algorithm needs to be employed that is capable of dealing with these conditions. These motion planning algorithms are often applied in conjunction with a local planning method that ensures that the trajectories generated by the planning algorithm adhere to the dynamic constraint of the UAV. The goal of this project is to find a motion planning algorithm that will provide safe flight for a rotary-wing UAV from an initial to a goal state in environments where dynamic obstacles exist. The motion planning algorithm implemented is based on the rapidly-exploring random tree (RRT) that is altered for dynamic environments and is called the real-time optimal RRT (RT-RRT*). Important changes were made to the RT-RRT* to increase the performance of the algorithm. The motion planning algorithm should adhere to the constraint of the vehicle, which is solved by employing a generic local planning method that make use of geometric-based motion primitives to construct trajectories that adhere to the constraints of the vehicle. The complete motion planning algorithm is tested thoroughly in various simulated environments and the performance of the algorithm is analysed. The motion planning algorithm was proven to be effective in sparse environments but struggled in more cluttered environments with multiple dynamic obstacles. Obstacle estimation was then implemented to try and improve the motion planning algorithm for cluttered environments, which proved to be an effective solution for avoiding dynamic obstacles. The trajectories generated by the motion planning algorithm was given to a realistic vehicle model to verify if the rotary-wing UAV will be able to accurately follow the generated trajectory. The vehicle model and controllers was previously designed and verified to be accurate during practical flight tests, which means that this model is an accurate representation of a real world vehicle.
AFRIKAANSE OPSOMMING: Vir 'n outonome onbemande lugvoertuig om veilig te navigeer in dinamiese omgewings, word dit vereis dat 'n konflikvry trajek tussen twee voertuig toestande vinnig en effektief beplan moet word. Meeste trajek beplannings metodes van vandag, is ontwerp om beplanning te doen in statiese omgewings en kan nie reageer op veranderinge wat in die omgewing kan plaasvind nie. Dus, vir n Onbemande lugvoertuig om veilig te navigeer deur dinamiese omgewings, is dit nodig om 'n bewegingsbeplannings algoritme te ontwerp wat hierdie kondisies kan hanteers. Beplannings algoritmes word dikwels in kombinasie met plaaslike beplannings metodes geimplimenteer sodat die trajekte wat gegenereer word, voldoen aan die dinamise beperkings van die voertuig. Die doel van hierdie projek is om 'n bewegingsbeplannings algoritme te ontwerp wat sal sorg dat 'n roterende vleuel Onbemande lugvoertuig veilig kan beweeg vanaf 'n aanvaklike tot 'n doel posisie deur 'n omgewing met dinamiese voorwerpe. Die bewegingsbeplannings algoritme wat geïmplementeer is, is gebaseer op die "Rapidly exploring Random Tree" (RRT) wat verander is om dinamiese omgewings te hanteer en word die "Real-Time RRT-star" (RTRRT*) genoem. Belangrike veranderinge was dan gemaak op die RT-RRT* om die uitvoering daarvan te verbeter. Die finale beplannings algoritme moet wel voldoen aan die dinamiese beperkings van die voertuig, wat opgelos is deur die implimenteering van 'n generiese plaaslike beplannings metode, wat gebruik maak van geometriese bewegins-primitiewe om 'n trajek te skep wat by die beperkings van die voertuig hou. Die finale beplannings algoritme word dan deeglik getoets in verskeie gesimuleerde omgewings sodat die uitvoer daarvan geanaliseer kan word. Dit was bewys dat die bewegingsbeplannings algoritme effektief werk vir omgewings met min voorwerpe, maar sukkel in besige omgewings waar daar verskeie dinamiese voorwerpe is. Hindernisberaming was toegepas om die algoritme te verbeter in besige omgewings en was bewys dat dit 'n effektiewe oplossing is om dinamiese voorwerpe te vermy. Die trajekte wat geskep word deur die beplannings algoritme was aan 'n realistiese voertuig model gestuur om te variieer dat die roterende vleuel Onbemande lugvoertuig hierdie trajekte akkuraat kan uitvoer. Die voertuig model en beheerders was voorheen ontwerp en getoets tydens praktiese vlug en is dus 'n geldige voorstelling van 'n regte wêreld voertuig.
Description
Thesis (MEng)--Stellenbosch University, 2020.
Keywords
Guidance systems (Flight), Trajectory optimization, Drone aircraft, UCTD, Airways -- Planning
Citation