Vizualno-inercijska odometrija zasnovana na optimizaciji faktorskih grafova

master's thesis

• Title (english): Visual-inertial odometry based on factor graph optimization
• Author: Karlo Koledić
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Marković (predsjednik povjerenstva)
• Committee member: Ivan Petrović (član povjerenstva)
• Committee member: Marija Seder (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

Lokalizacija autonomnih mobilnih platformi je ključna u današnjem svijetu. Mnoga područja poput robotike, autonomnih vozila, virtualne stvarnosti te bespilotnih letjelica zahtijevaju brzu i točnu estimaciju trenutne pozicije. U ovom radu se vrši teoretska i praktična analiza nove metode vizualno-inercijske odometrije baziranoj na preintegracijskoj teoriji na manifoldu koja uzima u obzir manifoldsku karakteristiku specijalne euklidske grupe SE(3). Odometrija je estimirana pomoću iSAM2 algoritma koji se bazira na grafičkom probabilističkom modelu faktorskih grafova. Metoda je testirana na EuRoC skupu podataka te na vlastitim sekvencama snimljenim pomoću dva različita senzora. Nadalje, u radu se testira i poboljšava neuronska mreža za estimaciju odometrije kako bi se istražili doprinosi dubokog učenja u estimaciji odometrije.

Abstract (english)

Localization of autonomous mobile platforms is essential in an increasingly autonomous world. Many applications such as robotics, autonomous vehicles, virtual reality and unmanned aerial vehicles require accurate and fast estimation of current position in the world. This work explores, both theoretically and experimentally, a novel Visual-Inertial Odometry method based on On-Manifold Preintegration Theory which correctly addresses manifold structure of Special Euclidian Group SE(3). Odometry is estimated with iSAM2 algorithm, which is based on factor graph probabilistic graphical model. Method is tested on EuRoC dataset and proprietary dataset recorded with two sensor modules. Additionally, in order to explore benefits of deep learning in odometry estimation, a novel neural network architecture is tested and improved.

• Keywords: odometrija; faktorski graf; iSAM2; manifold; Kimera; EuRoC; duboko učenje; Monodepth2
• Keywords (english): odometry; factor graph; iSAM2; manifold; Kimera; EuRoC; deep learning; Monodepth2
• Language: english
• URN:NBN: urn:nbn:hr:168:233476
• Study programme: Title: Electrical Engineering and Information Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije (magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 12:41:33