Search

Self-driving vehicles rely on multimodal motion forecasts to effectively interact with their environment and plan safe maneuvers. We introduce SceneMotion, an attention-based model for forecasting scene-wide motion modes of multiple traffic agents. Our model transforms local agent-centric embeddings into scene-wide forecasts using a novel latent context module. This module learns a scene-wide latent space from multiple agent-centric embeddings, enabling joint forecasting and interaction modeling. The competitive performance in the Waymo Open Interaction Prediction Challenge demonstrates the effectiveness of our approach. Moreover, we cluster future waypoints in time and space to quantify the interaction between agents. We merge all modes and analyze each mode independently to determine which clusters are resolved through interaction or result in conflict. Our implementation is available at: https://github.com/kit-mrt/future-motion, Comment: 7 pages, 3 figures, ITSC 2024; v2: added details about waypoint clustering

In this paper, we study a nonlinear cointegration-type model of the form \(Z_t = f_0(X_t) + W_t\) where \(f_0\) is a monotone function and \(X_t\) is a Harris recurrent Markov chain. We use a nonparametric Least Square Estimator to locally estimate \(f_0\), and under mild conditions, we show its strong consistency and obtain its rate of convergence. New results (of the Glivenko-Cantelli type) for localized null recurrent Markov chains are also proved.

It is the purpose of this paper to investigate the issue of estimating the regularity index $\beta>0$ of a discrete heavy-tailed r.v. $S$, \textit{i.e.} a r.v. $S$ valued in $\mathbb{N}^*$ such that $\mathbb{P}(S>n)=L(n)\cdot n^{-\beta}$ for all $n\geq 1$, where $L:\mathbb{R}^*_+\to \mathbb{R}_+$ is a slowly varying function. As a first go, we consider the situation where inference is based on independent copies $S_1,\; \ldots,\; S_n$ of the generic variable $S$. Just like the popular Hill estimator in the continuous heavy-tail situation, the estimator $\widehat{\beta}$ we propose can be derived by means of a suitable reformulation of the regularly varying condition, replacing $S$'s survivor function by its empirical counterpart. Under mild assumptions, a non-asymptotic bound for the deviation between $\widehat{\beta}$ and $\beta$ is established, as well as limit results (consistency and asymptotic normality). Beyond the i.i.d. case, the inference method proposed is extended to the estimation of the regularity index of a regenerative $\beta$-null recurrent Markov chain. Since the parameter $\beta$ can be then viewed as the tail index of the (regularly varying) distribution of the return time of the chain $X$ to any (pseudo-) regenerative set, in this case, the estimator is constructed from the successive regeneration times. Because the durations between consecutive regeneration times are asymptotically independent, we can prove that the consistency of the estimator promoted is preserved. In addition to the theoretical analysis carried out, simulation results provide empirical evidence of the relevance of the inference technique proposed.

Two regeneration-based bootstrap methods, namely, the \textit{Regeneration based-bootstrap} \cite{AthreyaFuh1992, Somnat-1993} and the \textit{Regenerative Block bootstrap} \cite{Bertail2006} are shown to be valid for the problem of estimating the integral of a function with respect to the invariant measure in a $\beta$-null recurrent Markov chain with an accessible atom. An extension of the Central Limit Theorem for randomly indexed sequences is also presented.

The main purpose of this article to extend the notion of statistical depth to the case of sample paths of a Markov chain. Initially introduced to define a center-outward ordering of points in the support of a multivariate distribution, depth functions permit to generalize the notions of quantiles and (signed) ranks for observations in $\mathbb{R}^d$ with $d>1$, as well as statistical procedures based on such quantities. In this paper, overcoming the lack of natural order on the torus composed of all possible trajectories of finite length, we develop a general theoretical framework for evaluating the depth of a Markov sample path and recovering it statistically from an estimate of its transition probability with (non-) asymptotic guarantees. We also detail some of its numerous applications, focusing particularly on anomaly detection, a key task in various fields involving the analysis of (supposedly) Markov time-series (\textit{e.g.} health monitoring of complex infrastructures, security). Beyond the description of the methodology promoted and the statistical analysis carried out to guarantee its validity, numerical experiments are displayed, providing strong empirical evidence of the relevance of the novel concept we introduce here to quantify the degree of abnormality of Markov path sequences of variable length.

We demonstrated the potential of the fractional dimensional approach to understand exciton parameters in the exemplary atomically thin semiconductor material, a monolayer of WS$_2$. This approach has proved to be successful in finding the exciton binding energy and quasiparticle bandgap for the WS$_2$ monolayer in varying dielectric environments. A tuning of the quasiparticle bandgap and binding energy by 141 meV and 188 meV, respectively, has been achieved by varying the dielectric of the environment from 1.52 to 8.1. The approach is justified by comparing the changes in the binding energy with the computational results from the Quantum Electrostatic Heterostructures model. The fractional dimension found through the excitonic Rydberg series is close to 2.8 for WS$_2$ monolayer in all different dielectric surroundings. Thus, this approach provides a rapid and robust method for determining the binding energy of excitons in 2D semiconductors independent of the particular dielectric environment.

We present JointMotion, a self-supervised pre-training method for joint motion prediction in self-driving vehicles. Our method jointly optimizes a scene-level objective connecting motion and environments, and an instance-level objective to refine learned representations. Scene-level representations are learned via non-contrastive similarity learning of past motion sequences and environment context. At the instance level, we use masked autoencoding to refine multimodal polyline representations. We complement this with an adaptive pre-training decoder that enables JointMotion to generalize across different environment representations, fusion mechanisms, and dataset characteristics. Notably, our method reduces the joint final displacement error of Wayformer, HPTR, and Scene Transformer models by 3\%, 8\%, and 12\%, respectively; and enables transfer learning between the Waymo Open Motion and the Argoverse 2 Motion Forecasting datasets. Code: https://github.com/kit-mrt/future-motion, Comment: CoRL'24 camera-ready

Any sports competition needs a timetable, specifying when and where teams meet each other. The recent International Timetabling Competition (ITC2021) on sports timetabling showed that, although it is possible to develop general algorithms, the performance of each algorithm varies considerably over the problem instances. This paper provides an instance space analysis for sports timetabling, resulting in powerful insights into the strengths and weaknesses of eight state-of-the-art algorithms. Based on machine learning techniques, we propose an algorithm selection system that predicts which algorithm is likely to perform best when given the characteristics of a sports timetabling problem instance. Furthermore, we identify which characteristics are important in making that prediction, providing insights in the performance of the algorithms, and suggestions to further improve them. Finally, we assess the empirical hardness of the instances. Our results are based on large computational experiments involving about 50 years of CPU time on more than 500 newly generated problem instances., Comment: This is the peer-reviewed author-version of https://doi.org/10.1016/j.ejor.2024.06.005, published in the European Journal of Operational Research. Copyright 2024. This manuscript version is made available under the LCC-BY-NC-ND 4.0 license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

We present a comprehensive study on the structure and optical properties of Mn-and Co-doped ZnO samples prepared via solid-state reaction method with different dopant concentrations and atmospheres. The samples were structural and chemically characterized via X-ray diffraction, scanning electron microscopy, energy-dispersive X-ray spectroscopy, and X-ray excited photoelectron spectroscopy. The optical characterization was performed via Raman, photoluminescence, and diffuse photoreflectance spectroscopies. Emphasis was done on the studies of their vibrational properties. The structural data confirm the incorporation of Mn and Co ions into the wurtzite ZnO lattice. It is demonstrated that the usual observed additional bands in the Raman spectrum of transitional metal (TM) doped ZnO are related to structural damage, deriving from the doping process, and surface effects. The promoted surface optical phonons (SOP) are of Fr\"ohlich character and, together with the longitudinal optical (LO) polar phonons, are directly dependent on the ZnO electronic structure. The enhancement of SOP and LO modes with TM-doping is explained in terms of nonhomogeneous doping, with the dopants concentrating mainly on the surface of grains, and a resonance effect due to the decrease of the ZnO bandgap promoted by the introduction of the 3d TM levels within the ZnO bandgap. We also discuss the origin of the controversial vibrational mode commonly observed in the Mn-doped ZnO system. It is stated that the observation of the analyzed vibrational properties is a signature of substitutional doping of the ZnO structure with tuning of ZnO optical absorption into the visible range of the electromagnetic spectrum.

Industry 4.0 applications impose the challenging demand of delivering packets with bounded latencies via a wireless network. This is further complicated if the network is not dedicated to the time critical application. In this paper we use network calculus analysis to derive closed form expressions of latency bounds for time critical traffic when 802.11 Target Wake Time (TWT) and 802.1Qbv work together in a shared 802.11 network., Comment: 3 pages, 3 figures, workshop submission

We introduce RedMotion, a transformer model for motion prediction in self-driving vehicles that learns environment representations via redundancy reduction. Our first type of redundancy reduction is induced by an internal transformer decoder and reduces a variable-sized set of local road environment tokens, representing road graphs and agent data, to a fixed-sized global embedding. The second type of redundancy reduction is obtained by self-supervised learning and applies the redundancy reduction principle to embeddings generated from augmented views of road environments. Our experiments reveal that our representation learning approach outperforms PreTraM, Traj-MAE, and GraphDINO in a semi-supervised setting. Moreover, RedMotion achieves competitive results compared to HPTR or MTR++ in the Waymo Motion Prediction Challenge. Our open-source implementation is available at: https://github.com/kit-mrt/future-motion, Comment: 17 pages, 8 figures; v2: focus on transformer model; v3: TMLR camera-ready

A key advantage of utilizing van der Waals materials as defect-hosting platforms for quantum applications is the controllable proximity of the defect to the surface or the substrate for improved light extraction, enhanced coupling with photonic elements, or more sensitive metrology. However, this aspect results in a significant challenge for defect identification and characterization, as the defect's optoelectronic properties depend on the specifics of the atomic environment. Here we explore the mechanisms by which the environment can influence the properties of carbon impurity centres in hexagonal boron nitride (hBN). We compare the optical and electronic properties of such defects between bulk-like and few-layer films, showing alteration of the zero-phonon line energies, modifications to their phonon sidebands, and enhancements of their inhomogeneous broadenings. To disentangle the various mechanisms responsible for these changes, including the atomic structure, electronic wavefunctions, and dielectric screening environment of the defect center, we combine ab-initio calculations based on a density-functional theory with a quantum embedding approach. By studying a variety of carbon-based defects embedded in monolayer and bulk hBN, we demonstrate that the dominant effect of the change in the environment is the screening of the density-density Coulomb interactions within and between the defect orbitals. Our comparative analysis of the experimental and theoretical findings paves the way for improved identification of defects in low-dimensional materials and the development of atomic scale sensors of dielectric environments., Comment: 38 pages, 7 figures

Last years we have witnessed the uprising of location based applications, which depend on the devices ability to accurately obtain their position. IEEE 802.11, foretelling the need for such applications, started the IEEE 802.11az work on Next Generation Positioning. Although this standard provides positioning enhancements for sub-6GHz and mmWave bands, high accuracy in the order of centimeters can only be obtained in the latter band, thanks to the beamforming information available at mmWave operation. This work presents a detailed analysis on the new techniques provided by IEEE 802.11az for enhanced secured positioning in the mmWave band, assessing them through experimentation., Comment: 8 pages, 6 figures, magazine submission

Augmented Reality (AR) applications are becoming more mainstream, with successful examples in the mobile environment like Pokemon GO. Current malicious techniques can exploit these environments' immersive and mixed nature (physical-virtual) to trick users into providing more personal information, i.e., dark patterns. Dark patterns are deceiving techniques (e.g., interface tricks) designed to influence individuals' behavioural decisions. However, there are few studies regarding dark patterns' potential issues in AR environments. In this work, using scenario construction to build our prototypes, we investigate the potential future approaches that dark patterns can have. We use VR mockups in our user study to analyze the effects of dark patterns in AR. Our study indicates that dark patterns are effective in immersive scenarios, and the use of novel techniques such as `haptic grabbing' to drag participants' attention can influence their movements. Finally, we discuss the impact of such malicious techniques and what techniques can mitigate them.

We explored a metal-insulator-metasurface structure exhibiting dual-band absorption in the visible and infrared regions with almost perfect absorbance. We demonstrated that the existence of the dual peak absorbance is due to the excitation of propagating surface plasmon and localized surface plasmon mode. We further utilized the excited degenerate propagating surface plasmon to enhance the fluorescence emission of an emitter located on top of the metasurface. This was accomplished by engineering a metasurface that excites a degenerate mode with resonance absorption bands close to the emission wavelength of fluorescent material. This condition helps to increase the rate of excitation and emission of an emitter by generating additional electron configurations when coupled to such metal-insulator-metasurface structures. This approach offers relevant potential in radiative engineering techniques., Comment: 8 pages, 4 figures

Human habitation across multiple planets requires communication and social connection between planets. When the infrastructure of a deep space network becomes mature, immersive cyberspace, known as the Metaverse, can exchange diversified user data and host multitudinous virtual worlds. Nevertheless, such immersive cyberspace unavoidably encounters latency in minutes, and thus operates in a turn-taking manner. This Blue Sky paper illustrates a vision of an interplanetary Metaverse that connects Earthian and Martian users in a turn-based Metaverse. Accordingly, we briefly discuss several grand challenges to catalyze research initiatives for the `Digital Big Bang' on Mars., Comment: Accepted Paper in ACM ICMI 2022 (Blue Sky Track)

One-dimensional optical waveguiding is revisited using the electromagnetic deduction of Fresnel formulas relating the incident, reflected, and transmitted waves on the abrupt interface between two different optical media. Throughout the paper, different optical configurations are analyzed from this perspective revealing that, in all cases, those solutions corresponding to waveguiding configurations share common features. These shared characteristics are: firstly, the guided modes always emerge as singular solutions of the associate coherent optical scattering problem and, secondly, the resonance condition that must be satisfied corresponds to the pole of the associated Fresnel Coefficients., Comment: 17 pages, 8 figures

The extinction law from $0.9$ to $8$ microns in the inner $3\times3$ deg$^2$ of the Milky Way is measured using data from VISTA Variables in the Via Lactea, GLIMPSE and WISE. Absolute extinction ratios are found by requiring that the observed red clump density peaks at the GRAVITY collaboration distance to the Galactic centre. When combined with selective extinction ratios measured from the bulge giant colour-colour diagrams, we find an extinction law of $A_Z:A_Y:A_J:A_H:A_{K_s}:A_{W1}:A_{[3.6]}:A_{[4.5]}:A_{W2}:A_{[5.8]}:A_{[8.0]} =7.19(0.30):5.11(0.20):3.23(0.11):1.77(0.04):1:0.54(0.02):0.46(0.03):0.34(0.03):0.32(0.03):0.24(0.04):0.28(0.03)$ valid for low extinctions where non-linearities are unimportant. These results imply an extinction law from the Rayleigh Jeans colour excess method (RJCE) of $A_{K_s}=0.677(H-[4.5]-0.188)$. We find little evidence for significant selective extinction ratio variation over the inspected region (around $5\%$). Assuming the absolute extinction ratios do not vary across the inspected region gives an independent measurement of the absolute $K_s$ magnitude of the red clump at the Galactic Centre of $(-1.61\pm0.07)\,\mathrm{mag}$. This is very similar to the value measured for solar neighbourhood red clump stars giving confidence in the use of red clump stars as standard candles across the Galaxy. As part of our analysis, we inspect the completeness of PSF photometry from the VVV survey using artificial star tests, finding $90\%$ completeness at $K_s\approx16 \,(17)$ in high (low) density regions and good agreement with the number counts with respect to the GALACTICNUCLEUS and DECAPS catalogues over small regions of the survey., Comment: 18 pages, 16 figures, accepted for publication in MNRAS

(Abridged) We have used Gaia EDR3 data, together with spectra of a large sample of luminous stars in the field surrounding Westerlund 1, to explore the extent of the cluster. We carry out a non-parametric analysis of proper motions and membership determination. We investigate the reddening and proper motions of several dozen OB stars and red supergiants < 1 deg away from Wd 1. We identify a population of kinematic members of Wd 1 that largely includes the known spectroscopic members. From their EDR3 parallaxes, we derive a distance to the cluster of $4.23^{+0.23}_{-0.21}$kpc. Extinction in this direction increases by a large amount around 2.8 kpc, due to dark clouds associated to the Scutum-Crux arm. As a consequence, we hardly see any stars at distances comparable to that of the cluster. The proper motions of Wd 1, however, are very similar to those of stars in the field surrounding it, but distinct. We find a second, astrometrically well-defined population in the foreground ($d\approx2\:$kpc), likely connected to the possible open cluster BH197. Wd 1 is very elongated, an effect not driven by the very heavy extinction to the East and South. We find a low-density halo extending up to 10' from the cluster centre, mainly in the NW quadrant. A few OB stars at larger distances from the cluster, most notably the LBV MN48, share its proper motions, suggesting that Wd 1 has little or no peculiar motion with respect to the field population of the Norma arm. However, we are unable to find any red supergiant that could belong to an extended population related to the cluster, although we observe several dozen such objects in the foreground. We find a substantial population of luminous OB members obscured by several more magnitudes of extinction than most known members. These objects, mostly located in the central region of the cluster, increase the population of OB supergiants by about 25%., Comment: Accepted for publication in Astronomy & Astrophysics. Figures have been converted to PNG and are of lower quality than in the version accepted. Three tables in the appendices will be published electronically; only table C1 is given here. 25 pages (including appendices),23 figures. V2 incorporates style and language edits and some comments