High-dimensional trajectory optimization for robot motion planning based on Gaussian processes

doctoral thesis

• Title (croatian): Optimizacija visokodimenzijskih trajektorija za planiranje gibanja robota zasnovana na Gaussovim procesima
• Author: Luka Petrović
• Mentor: Ivan Marković (mentor)
• Committee member: Stjepan Bogdan (predsjednik povjerenstva)
• Committee member: Marija Seder (član povjerenstva)
• Committee member: Marko Švaco (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Control and Computer Engineering); Zagreb
• Defense date and country: 2022-07-18, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

Motion planning is one of the fundamental tasks in robotics with the goal of producing feasible robot trajectories in configuration space that reach a desired goal optimally according to a given criterion. The principal purpose of this thesis is the development of novel high-dimensional motion planning methods that enable safe and efficient robot operation in real-world environments. We rely on trajectory optimization in order to produce robot trajectories that are smooth, collision-free and satisfy arbitrary task-dependent criteria. A common theme throughout this thesis is reliance on continuous-time Gaussian processes as trajectory representations that underpin efficient collision-checking and generate smooth robot trajectories. The thesis presents three scientific contributions: 1 A method for stochastic high-dimensional robot trajectory optimization based on cross-entropy and heteroscedastic Gaussian processes. 2 A high-dimensional trajectory optimization method based on stochastic gradient estimation using mixtures of Gaussian processes that enables inclusion of multiple complex criteria. 3 An incremental factor graph inference method for simultaneous moving object tracking and robot motion planning based on Gaussian processes.

Abstract (croatian)

Planiranje gibanja predstavlja jedan od temeljnih problema u robotici s ciljem generiranja ostvarivih robotskih trajektorija koje uspješno izvršavaju dani zadatak. Temeljni je cilj ove disertacije razvoj novih visokodimenzijskih algoritama planiranja gibanja koji omogućuju sigurno i učinkovito djelovanje robota u stvarnim, dinamičkim okruženjima. U radu se oslanjamo na metode optimizacije trajektorija kako bismo generirali glatke putanje robota koje nisu u sudaru s okolinom i zadovoljavaju proizvoljne optimizacijske kriterije. Oslanjamo se i na kontinuirane Gaussove procese za predstavljanje trajektorije čime je omogućena efikasna provjera sudara i postizanje visoke brzine izvođenja navedenih metoda. Tri su znanstvena doprinosa disertacije: 1 Metoda za stohastičku optimizaciju visokodimenzijske trajektorije robota zasnovana na unakrsnoj entropiji i heteroskedastičnim Gaussovim procesima. 2 Metoda optimizacije visokodimenzijskih trajektorija zasnovana na stohastičkoj estimaciji gradijenta korištenjem mješavine Gaussovih procesa koja omogućuje uključivanje više složenih kriterija. 3 Inkrementalna metoda zaključivanja na faktor grafu za istovremeno praćenje gibajućih objekata i planiranje gibanja robota zasnovana na Gaussovim procesima.

• Keywords: high-dimensional motion planning; trajectory optimization; Gaussian processes; collision avoidance; singularity avoidance; stochastic optimization; cross-entropy optimization; mixtures of Gaussian processes; robotic arm; mobile manipulator
• Keywords (croatian): visokodimenzijsko planiranje gibanja; optimizacija trajektorija; Gaussovi procesi; izbjegavanje sudara; izbjegavanje singulariteta; stohastička optimizacija; optimizacija unakrsne entropije; mješavine Gaussovih procesa; robotska roka; mobilni manipulator
• Language: english
• URN:NBN: urn:nbn:hr:168:262919
• Promotion: 2022
• Study programme: Title: Computer Science; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti (Doktor znanosti)
• Type of resource: Text
• Extent: xvi, 135 str. : ilustr. u boji ; 30 cm.
• File origin: Born digital
• Access conditions: Access restricted to students and staff of home institution
• Created on: 2022-09-12 08:22:07