Portfolio item number 1
Short description of portfolio item number 1
Posts by Collection
portfolio
Portfolio item number 2
Short description of portfolio item number 2
publications
RNN-based observability analysis for magnetometer-free sparse inertial motion tracking
Published in 2022 25th International Conference on Information Fusion, 2022
We present an RNN-based method for assessing the observability of relative pose in sparse, magnetometer-free inertial motion tracking (IMT) systems, demonstrating that observability and tracking accuracy depend on the kinematic structure, thus advancing reliable and cost-effective IMT solutions for complex kinematic chains.
Neural ODEs for Data-Driven Automatic Self-Design of Finite-Time Output Feedback Control for Unknown Nonlinear Dynamics
Published in IEEE Control System Letters, 2023
ANODEC is a data-efficient, learning-based method for automatically designing output feedback controllers for finite-time reference tracking in unknown nonlinear systems, achieving superior accuracy over baselines across diverse reference signals and system dynamics.
Plug-and-Play Sparse Inertial Motion Tracking With Sim-to-Real Transfer
Published in IEEE Sensors Letters, 2023
We propose a magnetometer-free, sparse IMU-based inertial motion tracking (IMT) method using a recurrent neural network observer (RNNo) trained on simulated data, achieving plug-and-play generalizability and robust tracking with a <4 degree error in complex kinematic chains.
A Soft Robotic System Automatically Learns Precise Agile Motions Without Model Information
Published in 2024 IEEE/RSJ international conference on intelligent robots and systems, 2024
This work demonstrates that the Automatic Neural ODE Control (ANODEC) method enables practical and efficient control of pneumatic soft robots (SRs) with hysteresis effects, achieving agile, non-repetitive reference tracking from only 30 seconds of input-output data and outperforming manually tuned PID controllers, thereby advancing the feasibility of data-driven, low-expertise SR control.
Recurrent Inertial Graph-Based Estimator (RING): A Single Pluripotent Inertial Motion Tracking Solution
Published in Transactions on Machine Learning Research, 2024
This paper presents RING, a novel machine learning-based, plug-and-play method for Inertial Motion Tracking (IMT) that eliminates the need for expert knowledge by employing a decentralized network of recurrent neural networks, enabling broad applicability, zero-shot generalization from simulation to experimental data, and high performance across diverse IMT challenges, including magnetometer-free and sparse sensing setups.
talks
teaching
Tutor for Inertial Sensor Fusion
Graduate course, FAU Erlangen-Nürnberg, Department Artificial Intelligence in Biomedical Engineering, 2021
This course is concerned with inertial sensor technologies and sensor fusion methods for motion tracking of aerial/ground/water vehicles, robotic systems and human body segments.
Tutor for Artificial Motor Learning
Graduate course, FAU Erlangen-Nürnberg, Department Artificial Intelligence in Biomedical Engineering, 2022
This course is concerned with methods of artificial intelligence that enable biomimetic motor learning in intelligent systems.
Lecturer for Introduction to Explainable Machine Learning
Graduate course, FAU Erlangen-Nürnberg, Department Artificial Intelligence in Biomedical Engineering, 2023
This course gives an introduction to explainable and interpretable methods and approaches in machine learning. We discuss prominent concepts in explainable machine learning, analyze and compare their potential and shortcomings, and apply them to example problems.