I am currently a Postdoctoral Fellow working with Stefan Weiss at the Control of Networked Systems, Austria. I obtained my Ph.D. in Computer Science at the University of Klagenfurt, Austria. Prior to this, I completed both a B.Sc. and an M.Sc. in Hardware Software Design and Embeeded Systems Design at the Univeristy of Applied Sciences Hagenberg, Austria.

My research is primarily concerned with robust localization in swarms of heterogeneous mobile agents with multiple and time-varying sensing modalities focusing on the design, modeling, simulation and evaluation. In my Ph.D. thesis, supervised by Stefan Weiss, we investigate the issue of robust localization in swarms of heterogeneous mobile agents with multiple and time-varying sensing modalities. Our focus is the development of filter-based and decoupled estimators under the assumption that agents possess communication and processing capabilities. My interest originates from my stay at the Austrian Institute of Technology (AIT) with the Vision, Automation, and Control department, where I worked, e.g., on a real-time stereo matching algorithm (CUDA) for an embedded GPU accelerated platform, an a drive assistant system that detects obstacle behind a wheel loader, and on mobile visual SLAM algorithm. My interests are Ultra-wideband Inertial Navigation, Collaborative and Modular State Estimation, Computer Vision, Embedded Autonomous Systems and Robotics.

Roland Jung, Ph.D.

Postdoctoral Fellow

Control of Networked Systems
University of Klagenfurt
Austria

home
Map
roland.jung@ieee.org
ResearchGate
jungr-ait
LinkedIn
Google Scholar

Selected Research Projects


Networked Autonomous Aerial Vehicles (KPK NAV), 2017-2021
Was a doctoral program of University of Klagenfurt for four doctoral students — Agata Barciś, Michał Barciś, Roland Jung and Petra Maždin — focusing on the topics of information distribution, decentralised synchronisation and coordination, and collaborative state estimation. The aim of the final demonstrator was to carry out a mission involving four drones — from task distribution, coalition building, collision avoidance, optimised information distribution, constellations for synchronised image capture to the generation of a 3D model of the captured objects.
FFG RoMInG, 2022-2025
The FFG Bridge project RoMInG aimed to develop a unified probabilistic framework for collaborative state estimation and multi-sensor fusion for an autonomous UAV-based bridge inspection. While full real-world validation faced challenges due to sensor limitations, significant theoretical and software advancements were made, including the development of the DC-MMSF algorithm, a modular estimation framework (mmsf_ros), and a realistic multi-agent simulation environment. Investigations into alternative perception systems (UVDAR) and improved ranging protocols (TDMA-based UWB) supported the project’s direction. Despite difficulties in acquiring usable real-world data, the project generated valuable insights and disseminated findings through multiple publications and presentations. A novel and generic filter formulation termed Isolated Kalman Filter (IsoKF) was proposed and empirically validated for Decoupled-Input and Coupled-Output (DICO) systems, demonstrating stable and consistent estimates. This work establishes a unifying approach to decouple estimates across single agents and multi-agent swarms. Based on the IsoKF, the Distributed Collaborative Modular Multi-Sensor Fusion (DC-MMSF) algorithm for collaborative state estimation was proposed. A realistic multi-agent simulation environment was created in Gazebo using a 3D model of the Kleindürrenbachbrücke, supporting multiple UAVs with various sensor configurations. A new sensor plugin in Gazebo was implemented for meshed UWB ranging. Simulation results mirrored issues observed with real-world UVDAR data.
An alternative mutual perception system was evaluated in a collaboration with the Technical University of Prague. This led to the integration and evaluation of the UVDAR system (visual mutual perception utilizing blinking UV-LEDs). While UVDAR provided 6-DoF relative position and orientation estimates, its use for relative localization within ROMING was limited.
A dataset was recorded for the UVDAR evaluation, revealing inconsistencies and limitations in dynamic scenarios. The UVDAR system was found unsuitable for the project’s localization requirements. Therefore, a novel UWB ranging protocol was developed. The initial meshed UWB protocol’s limitations prompted the development of a dynamic TDMA-based double-sided two-way ranging (DS-TWR) protocol and a custom firmware for Qorvo DMW1001 modules was implemented, and antenna delay and anchor position calibration methods were investigated. The 3D BIM generation workflow utilizing LiDAR point clouds for 3D model reconstruction was evaluated using the realistic Gazebo simulation environment. Ring segments were recorded in simulation, and mesh extraction yielded accurate results. A dataset was recorded at the Kleindürrenbachbrücke using 4 UAVs with LiDAR and UWB sensors.

Selected Publications

PhD Thesis

Recursive Distributed Collaborative Aided Inertial Navigation
Roland Jung;
University of Klagenfurt, 2023
Journal Articles
Conference Publications
Workshop Contributions
Preprints
    Datasets

    Selected Videos



    Selected Frameworks and Tools


    ikf_lib
    This repository contains an Isolated Kalman Filtering framework, implementing the isolated Kalman filter (IKF) paradigm with support for out-of-sequence measurements. The primary motivation of the IKF paradigm is to decouple correlated estimates of filter instances by employing approximations proposed by Luft et al.

    mmsf_lib
    The mmsf_lib aims at modular sensor fusion and builds upon the ikf_lib. It defines different sensor model, estimates and sensor specific estimators (deriving from the ikf::IIsolatedKalmanfilter). A factory allows to instantiate/create sensor-specific sensor estimator objects, which are maintained in a SensorInstanceHandler which derives from the ikf::IsolatedKalmanfilterHandler. The local full-state is defined by the state-info, which allows to mapping between states name and the actual position/range in the vector. This is in particular important to access state element, e.g., to compute Jacobians, for plotting, or for debugging purposes. Conculding, this library implements actual sensor nodes (prediction and update models, i.e. the Jacobians and state definitions).

    mmsf_ros
    The mmsf_ros is a ROS1 wrapper for the mmsf_lib and allows to create bag files from a sensor suite with measurement data, which can be generated, e.g., in corresponding the MATLAB framework. It support collaborative state estimation (CSE) by processing measurements between to mmsf_ros nodes - so called inter-agent observations. The inter-agent communication is handled via ROS services. Agents are automatically fetching meta information from each other.

    Working Experience

    Junior Scientis at AIT, 03/2015–11/2017
    At the Austrian Institute of Technology (AIT) with the Vision, Automation, and Control department, supervised by Dr. Martin Humenberger.
    My responsibiliets were the implemenation and porting of an real-time capable stereo vision algorithm in CUDA for the NVIDIA Jetson Tegra TK1 platform. Based on that, I developed an assisting system aimed for a wheel loader, that identifies and tracks obstacles (humans) in the rear of the vehicle. I served in different other projects, e.g., in a visual airplain tracking and intersection detection pipeline. In my last project we were developing a mobile Visual-SLAM algorthim (IPhone 7) for indoor localization and relocalization based on pretrained envirnment through server-client architecture in cooperation with Insider Navigation.

    Summer Research Assistant at B&R, 03/2013–09/2013
    At B&R Industrial Automation with the Safetyand Automation department, supervised by Dr. Stefan Schönegger.
    My main focus was porting, evaluating the openPOWERLINK Stack on new FPGA development Kits, such as the Terasic DE1-SoC Development Kit.

    Plain Academic