arXiv subjects

Matthias Drüppel

Publications and source records attributed to Matthias Drüppel.

4 recordsLinked to original sources

arXiv2026

Tiny Neural Networks for Multi-Object Tracking in a Modular Kalman Framework

We present a modular, production-ready approach that integrates compact Neural Network (NN) into a Kalmanfilter-based Multi-Object Tracking (MOT) pipeline. We design three tiny task-specific networks to retain modularity, interpretability and eal-time suitability for embedded Automotive Driver Assistance Systems: (i) SPENT (Single-Prediction Network) - predicts per-track states and replaces heuristic motion models used by the Kalman Filter (KF). (ii) SANT (Single-Association Network) - assigns a single incoming sensor object to existing tracks, without relying on heuristic distance and association metrics. (iii) MANTa (Multi-Association Network) - jointly associates multiple sensor objects to multiple tracks in a single step. Each module has less than 50k trainable parameters. Furthermore, all three can be operated in real-time, are trained from tracking data, and expose modular interfaces so they can be integrated with standard Kalman-filter state updates and track management. This makes them drop-in compatible with many existing trackers. Modularity is ensured, as each network can be trained and evaluated independently of the others. Our evaluation on the KITTI tracking benchmark shows that SPENT reduces prediction RMSE by more than 50% compared to a standard Kalman filter, while SANT and MANTa achieve up to 95% assignment accuracy. These results demonstrate that small, task-specific neural modules can substantially improve tracking accuracy and robustness without sacrificing modularity, interpretability, or the real-time constraints required for automotive deployment.

cs.CV↗

arXiv2018

Thickness-Dependent Differential Reflectance Spectra of Monolayer and Few-Layer MoS2, MoSe2, WS2 and WSe2

The research field of two dimensional (2D) materials strongly relies on optical microscopy characterization tools to identify atomically thin materials and to determine their number of layers. Moreover, optical microscopy-based techniques opened the door to study the optical properties of these nanomaterials. We presented a comprehensive study of the differential reflectance spectra of 2D semiconducting transition metal dichalcogenides (TMDCs), MoS2, MoSe2, WS2, and WSe2, with thickness ranging from one layer up to six layers. We analyzed the thickness-dependent energy of the different excitonic features, indicating the change in the band structure of the different TMDC materials with the number of layers. Our work provided a route to employ differential reflectance spectroscopy for determining the number of layers of MoS2, MoSe2, WS2, and WSe2.

cond-mat.mtrl-sci↗

arXiv2017

Biaxial strain tuning of the optical properties of single-layer transition metal dichalcogenides

Since their discovery single-layer semiconducting transition metal dichalcogenides have attracted much attention thanks to their outstanding optical and mechanical properties. Strain engineering in these two-dimensional materials aims to tune their bandgap energy and to modify their optoelectronic properties by the application of external strain. In this paper we demonstrate that biaxial strain, both tensile and compressive, can be applied and released in a timescale of a few seconds in a reproducible way on transition metal dichalcogenides monolayers deposited on polymeric substrates. We can control the amount of biaxial strain applied by letting the substrate expand or compress. To do this we change the substrate temperature and choose materials with a large thermal expansion coefficient. After the investigation of the substrate-dependent strain transfer, we performed micro-differential spectroscopy of four transition metal dichalcogenides monolayers (MoS2, MoSe2, WS2, WSe2) under the application of biaxial strain and measured their optical properties. For tensile strain we observe a redshift of the bandgap that reaches a value as large as 95 meV/% in the case of single-layer WS2 deposited on polypropylene. The observed bandgap shifts as a function of substrate extension/compression follow the order MoSe2 < MoS2 < WSe2 < WS2. Theoretical calculations of these four materials under biaxial strain predict the same trend for the material-dependent rates of the shift and reproduce well the features observed in the measured reflectance spectra.

cond-mat.mes-hall↗

arXiv2016

Three-particle correlation from a Many-Body Perspective: Trions in a Carbon Nanotube

Trion states of three correlated particles (e.g., two electrons and one hole) are essential to understand the optical spectra of doped or gated nanostructures, like carbon nanotubes or transition-metal dichalcogenides. We develop a theoretical many-body description for such correlated states using an ab-initio approach. It can be regarded as an extension of the widely used $GW$ method and Bethe-Salpeter equation, thus allowing for a direct comparison with excitons. We apply this method to a semiconducting (8,0) carbon nanotube, and find that the lowest optically active trions are red-shifted by $\sim 130$ meV compared to the excitons, confirming experimental findings for similar tubes. Moreover, our method provides detailed insights in the physical nature of trion states. In the prototypical carbon nanotube we find a variety of different excitations, discuss the spectra, energy compositions, and correlated wave functions.

cond-mat.mes-hall↗