Fabian Buse

Fabian Buse

Auto-generated profile, based on publicly available information

Fabian Buse

Institute of System Dynamics and Control, German Aerospace Center (DLR), Wessling, Germany

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 October 2024 in Machine Learning: Science and Technology
Verified
Journal Article
Open access
KRKislaya Ravi
Kislaya Ravi
VFVladyslav Fediukov
Vladyslav Fediukov
FDFelix Dietrich
Felix Dietrich
TNTobias Neckel
Tobias Neckel
FBFabian Buse
Fabian Buse
HBHans-Joachim Bungartz
Hans-Joachim Bungartz
One of the main challenges in surrogate modeling is the limited availability of data due to resource constraints associated with computationally expensive simulations. Multi-fidelity methods provide a solution by chaining models in a hierarchy with increasing fidelity, associated with lower error, but increasing cost. In this paper, we compare different multi-fidelity methods employed in constructing Gaussian process surrogates for regression. Non-linear autoregressive methods in the existing literature are primarily confined to two-fidelity models, and we extend these methods to handle more than two levels of fidelity. Additionally, we propose enhancements for an existing method incorporating delay terms by introducing a structured kernel. We demonstrate the performance of these methods across various academic and real-world scenarios. Our findings reveal that multi-fidelity methods generally have a smaller prediction error for the same computational cost as compared to the single-fidelity method, although their effectiveness varies across different scenarios.

9 July 2025
Verified
Conference Paper
Open access
FBFabian Buse
Fabian Buse
RHRoman Holderried
Roman Holderried
SLSandra Lagabarre
Sandra Lagabarre
NMNaomi Murdoch
Naomi Murdoch
JSJuliane Skibbe
Juliane Skibbe
PVPierre Vernazza
Pierre Vernazza
IDEFIX is a surface rover developed by CNES and DLR as part of JAXA's Martian Moons eXploration (MMX) mission. Designed to explore the surface of Phobos, it is the first wheeled robotic system intended to operate in a milli-gravity environment [1]. The rover's mobility system is central to its ability to conduct scientific investigations and navigate safely across uncertain and variable terrain. The rover hardware is already completed and awaits its launch towards Phobos in 2026.Figure 1: IDEFIX in its flight configuration awaiting shipment to Japan.The rover carries four instruments: a set of two navigation cameras (NavCam), a set of two cameras designed to analyze the interaction between the wheels and regolith (WheelCams) [2], a Raman spectrometer (RAX) analyzing the surface's mineralogical composition, and a thermal mapper (MiniRAD) [1].To realize the rover's mobility, it brings five key contributors:Its locomotion system enables mobility by allowing the rover to move across the surface and adjust its orientation. Its Navigation cameras provide the visual information needed for onboard or ground-based navigation. Its wheel cameras provide crucial insights into the wheel regolith interaction, enabling the rover team to asses safe driving strategies. It has two autonomous onboard navigation systems, one developed by CNES and one by DLR, enabling the rover to undertake longer drives. Its attitude control system (SKA) provides absolute attitude information to ground and board tools and autonomously commands the rover to align itself toward an optimal charging orientation. The locomotion system [3] is driven by the requirement to aid in the rover's initial uprighting. Thus, it unfolds and refolds its legs to orient the rover from any initial orientation after landing on the surface towards its belly. This requirement leads to the design of individually actuated legs with independently driven wheels. The locomotion system must rely on skid steering. The wheels are designed to operate on even extremely soft regolith while providing controllability on a more rigid regolith. With limited knowledge of the regolith behavior on Phobos, no specific conditions were assumed during the system's design. Instead, it is designed to operate under a wide range of conditions. Three locomotion system functions are described below; they are key in operating the rover.DRIVE executes a skid steer driving based on a commanded distance and heading angle change. In addition to these base arguments, traction can be shifted between its front and rear wheels, as this may be required to suppress a wheely-like behavior due to too high traction on the rear wheels.INCHING realizes an inch-worm-like movement. Designed to traverse especially soft or steep passages. ALIGN changes the rover chassis orientation and height while suppressing longitudinal movements. This command allows the SKA [4] system to safely realign the rover towards the sun.The rover's position and abilities in its environment are reconstructed on-ground based on:camera images: local and regional terrain models, local terrain identification 3-axis estimation computed by the attitude control system and the locomotion position: full rover pose estimation locomotion system telemetry, foremost leg and wheel angles. Besides, the behavior will be analyzed by comparing expected and real movements to update movement rules and improve mobility planning [5].Moving the rover on Phobos only through ground operation limits mobility to short explicit movements that must be manually confirmed. Thus, this mode of operation will limit mobility efficiency to short distances.Both navigation systems [5,6] aim to extend this limitation by providing a function to give higher-level commands, so changing the given command from "drive 0.5m and turn 15°" to "go to position [155, 188] in the mission frame (about 15 m ahead) with orientation 30°". There are several differences here: the first will only be executed in an open-loop, while the latter is executed over multiple days, establishes a path along the way to avoid hazardous areas, is controlled in closed-loop manner, and ensures the rover reaches its goal with some accuracy. Further, both systems provide functions to ensure the rover's safety, e.g., monitoring effects like slippage and aborting when a threshold is exceeded.Figure 2: (left) Prototypix, IDEFIX's earthbound validation and testing platform. (right) IDEFIX in simulation.Multiple means are available to test, validate, and train IDEFIX's mobility, from a simulation dedicated to only locomotion to dynamic simulation coupled in a hardware-in-the-loop manner connecting a flight representative onboard computer running the flight software [7], see Figure 2 (right).Testing mobility with a physical rover is limited as the effects of the difference in gravity are severe. First, the locomotion system is not designed to operate in Earth's gravity. While this issue can be solved using a gravity offloading system, the effects on regolith cannot be correctly replicated. Thus, tests with a complete rover system are limited to qualitative verification and some training aspects, see Figure 2 (left).The IDEFIX rover's mobility systems are designed to address the challenges of operating in Phobos' milli-gravity environment. Our efforts aim to ensure safe, efficient mobility and maximize the scientific return of the MMX mission.[1] ULAMEC, Stephan, et al. Science objectives of the MMX rover. Acta Astronautica, 2023.[2] Murdoch N. et al. The WheelCams on the IDEFIX rover, Submitted to PEPS."[3] BARTHELMES, Stefan, et al. MMX rover locomotion subsystem-development and testing towards the flight model. In: 2022 IEEE Aerospace Conference, AERO 2022. IEEE, 2022.[4] LAGABARRE, Sandra, et al. Design of the MMX Rover Attitude Control System for Autonomous Power Supply. ESA GNC-ICATT 2023, 2023.[5] BUSE, Fabian, et al. Mobility on the Surface of Phobos for the MMX...