Showing 1–10 of 10 results for author: Brucker, M

Geometrically Inspired Kernel Machines for Collaborative Learning Beyond Gradient Descent

Authors: Mohit Kumar, Alexander Valentinitsch, Magdalena Fuchs, Mathias Brucker, Juliana Bowles, Adnan Husakovic, Ali Abbas, Bernhard A. Moser

Abstract: This paper develops a novel mathematical framework for collaborative learning by means of geometrically inspired kernel machines which includes statements on the bounds of generalisation and approximation errors, and sample complexity. For classification problems, this approach allows us to learn bounded geometric structures around given data points and hence solve the global model learning proble… ▽ More This paper develops a novel mathematical framework for collaborative learning by means of geometrically inspired kernel machines which includes statements on the bounds of generalisation and approximation errors, and sample complexity. For classification problems, this approach allows us to learn bounded geometric structures around given data points and hence solve the global model learning problem in an efficient way by exploiting convexity properties of the related optimisation problem in a Reproducing Kernel Hilbert Space (RKHS). In this way, we can reduce classification problems to determining the closest bounded geometric structure from a given data point. Further advantages that come with our solution is that our approach does not require clients to perform multiple epochs of local optimisation using stochastic gradient descent, nor require rounds of communication between client/server for optimising the global model. We highlight that numerous experiments have shown that the proposed method is a competitive alternative to the state-of-the-art. △ Less

Submitted 5 July, 2024; originally announced July 2024.

Photometry of the Didymos system across the DART impact apparition

Authors: Nicholas Moskovitz, Cristina Thomas, Petr Pravec, Tim Lister, Tom Polakis, David Osip, Theodore Kareta, Agata Rożek, Steven R. Chesley, Shantanu P. Naidu, Peter Scheirich, William Ryan, Eileen Ryan, Brian Skiff, Colin Snodgrass, Matthew M. Knight, Andrew S. Rivkin, Nancy L. Chabot, Vova Ayvazian, Irina Belskaya, Zouhair Benkhaldoun, Daniel N. Berteşteanu, Mariangela Bonavita, Terrence H. Bressi, Melissa J. Brucker , et al. (56 additional authors not shown)

Abstract: On 26 September 2022, the Double Asteroid Redirection Test (DART) spacecraft impacted Dimorphos, the satellite of binary near-Earth asteroid (65803) Didymos. This demonstrated the efficacy of a kinetic impactor for planetary defense by changing the orbital period of Dimorphos by 33 minutes (Thomas et al. 2023). Measuring the period change relied heavily on a coordinated campaign of lightcurve phot… ▽ More On 26 September 2022, the Double Asteroid Redirection Test (DART) spacecraft impacted Dimorphos, the satellite of binary near-Earth asteroid (65803) Didymos. This demonstrated the efficacy of a kinetic impactor for planetary defense by changing the orbital period of Dimorphos by 33 minutes (Thomas et al. 2023). Measuring the period change relied heavily on a coordinated campaign of lightcurve photometry designed to detect mutual events (occultations and eclipses) as a direct probe of the satellite's orbital period. A total of 28 telescopes contributed 224 individual lightcurves during the impact apparition from July 2022 to February 2023. We focus here on decomposable lightcurves, i.e. those from which mutual events could be extracted. We describe our process of lightcurve decomposition and use that to release the full data set for future analysis. We leverage these data to place constraints on the post-impact evolution of ejecta. The measured depths of mutual events relative to models showed that the ejecta became optically thin within the first ~1 day after impact, and then faded with a decay time of about 25 days. The bulk magnitude of the system showed that ejecta no longer contributed measurable brightness enhancement after about 20 days post-impact. This bulk photometric behavior was not well represented by an HG photometric model. An HG1G2 model did fit the data well across a wide range of phase angles. Lastly, we note the presence of an ejecta tail through at least March 2023. Its persistence implied ongoing escape of ejecta from the system many months after DART impact. △ Less

Submitted 3 November, 2023; originally announced November 2023.

Comments: 52 pages, 5 tables, 9 figures, accepted to PSJ

Local and Global Information in Obstacle Detection on Railway Tracks

Authors: Matthias Brucker, Andrei Cramariuc, Cornelius von Einem, Roland Siegwart, Cesar Cadena

Abstract: Reliable obstacle detection on railways could help prevent collisions that result in injuries and potentially damage or derail the train. Unfortunately, generic object detectors do not have enough classes to account for all possible scenarios, and datasets featuring objects on railways are challenging to obtain. We propose utilizing a shallow network to learn railway segmentation from normal railw… ▽ More Reliable obstacle detection on railways could help prevent collisions that result in injuries and potentially damage or derail the train. Unfortunately, generic object detectors do not have enough classes to account for all possible scenarios, and datasets featuring objects on railways are challenging to obtain. We propose utilizing a shallow network to learn railway segmentation from normal railway images. The limited receptive field of the network prevents overconfident predictions and allows the network to focus on the locally very distinct and repetitive patterns of the railway environment. Additionally, we explore the controlled inclusion of global information by learning to hallucinate obstacle-free images. We evaluate our method on a custom dataset featuring railway images with artificially augmented obstacles. Our proposed method outperforms other learning-based baseline methods. △ Less

Submitted 28 July, 2023; originally announced July 2023.

InfraDet3D: Multi-Modal 3D Object Detection based on Roadside Infrastructure Camera and LiDAR Sensors

Authors: Walter Zimmer, Joseph Birkner, Marcel Brucker, Huu Tung Nguyen, Stefan Petrovski, Bohan Wang, Alois C. Knoll

Abstract: Current multi-modal object detection approaches focus on the vehicle domain and are limited in the perception range and the processing capabilities. Roadside sensor units (RSUs) introduce a new domain for perception systems and leverage altitude to observe traffic. Cameras and LiDARs mounted on gantry bridges increase the perception range and produce a full digital twin of the traffic. In this wor… ▽ More Current multi-modal object detection approaches focus on the vehicle domain and are limited in the perception range and the processing capabilities. Roadside sensor units (RSUs) introduce a new domain for perception systems and leverage altitude to observe traffic. Cameras and LiDARs mounted on gantry bridges increase the perception range and produce a full digital twin of the traffic. In this work, we introduce InfraDet3D, a multi-modal 3D object detector for roadside infrastructure sensors. We fuse two LiDARs using early fusion and further incorporate detections from monocular cameras to increase the robustness and to detect small objects. Our monocular 3D detection module uses HD maps to ground object yaw hypotheses, improving the final perception results. The perception framework is deployed on a real-world intersection that is part of the A9 Test Stretch in Munich, Germany. We perform several ablation studies and experiments and show that fusing two LiDARs with two cameras leads to an improvement of +1.90 mAP compared to a camera-only solution. We evaluate our results on the A9 infrastructure dataset and achieve 68.48 mAP on the test set. The dataset and code will be available at https://a9-dataset.com to allow the research community to further improve the perception results and make autonomous driving safer. △ Less

Submitted 29 April, 2023; originally announced May 2023.

Recent formation and likely cometary activity of near-Earth asteroid pair 2019 PR2 -- 2019 QR6

Authors: Petr Fatka, Nicholas A. Moskovitz, Petr Pravec, Marco Micheli, Maxime Devogèle, Annika Gustafsson, Jay Kueny, Brian Skiff, Peter Kušnirák, Eric Christensen, Judit Ries, Melissa Brucker, Robert McMillan, Jeffrey Larsen, Ron Mastaler, Terry Bressi

Abstract: Asteroid pairs are genetically related asteroids that recently separated ($<$few million years), but still reside on similar heliocentric orbits. A few hundred of these systems have been identified, primarily in the asteroid main-belt. Here we studied a newly discovered pair of near-Earth objects (NEOs): 2019 PR2 and 2019 QR6. Based on broad-band photometry, we found these asteroids to be spectral… ▽ More Asteroid pairs are genetically related asteroids that recently separated ($<$few million years), but still reside on similar heliocentric orbits. A few hundred of these systems have been identified, primarily in the asteroid main-belt. Here we studied a newly discovered pair of near-Earth objects (NEOs): 2019 PR2 and 2019 QR6. Based on broad-band photometry, we found these asteroids to be spectrally similar to D-types, a type rare amongst NEOs. We recovered astrometric observations for both asteroids from the Catalina Sky Survey from 2005, which significantly improved their fitted orbits. With these refinements we ran backwards orbital integrations to study formation and evolutionary history. We found that neither a pure gravitational model nor a model with the Yarkovsky effect could explain their current orbits. We thus implemented two models of comet-like non-gravitational forces based on water or CO sublimation. The first model assumed quasi-continuous, comet-like activity after separation, which suggested a formation time of the asteroid pair $300^{+120}_{-70}$ years ago. The second model assumed short-term activity for up to one heliocentric orbit ($\sim$13.9 years) after separation, which suggested that the pair formed 272$\pm$7 years ago. Image stacks showed no activity for 2019~PR2 during its last perihelion passage. These results strongly argue for a common origin that makes these objects the youngest asteroid pair known to date. Questions remain regarding whether these objects derived from a parent comet or asteroid, and how activity may have evolved since their separation. △ Less

Submitted 2 February, 2022; v1 submitted 2 December, 2021; originally announced December 2021.

Complex Robotic Manipulation via Graph-Based Hindsight Goal Generation

Authors: Zhenshan Bing, Matthias Brucker, Fabrice O. Morin, Kai Huang, Alois Knoll

Abstract: Reinforcement learning algorithms such as hindsight experience replay (HER) and hindsight goal generation (HGG) have been able to solve challenging robotic manipulation tasks in multi-goal settings with sparse rewards. HER achieves its training success through hindsight replays of past experience with heuristic goals, but under-performs in challenging tasks in which goals are difficult to explor… ▽ More Reinforcement learning algorithms such as hindsight experience replay (HER) and hindsight goal generation (HGG) have been able to solve challenging robotic manipulation tasks in multi-goal settings with sparse rewards. HER achieves its training success through hindsight replays of past experience with heuristic goals, but under-performs in challenging tasks in which goals are difficult to explore. HGG enhances HER by selecting intermediate goals that are easy to achieve in the short term and promising to lead to target goals in the long term. This guided exploration makes HGG applicable to tasks in which target goals are far away from the object's initial position. However, HGG is not applicable to manipulation tasks with obstacles because the euclidean metric used for HGG is not an accurate distance metric in such environments. In this paper, we propose graph-based hindsight goal generation (G-HGG), an extension of HGG selecting hindsight goals based on shortest distances in an obstacle-avoiding graph, which is a discrete representation of the environment. We evaluated G-HGG on four challenging manipulation tasks with obstacles, where significant enhancements in both sample efficiency and overall success rate are shown over HGG and HER. Videos can be viewed at https://sites.google.com/view/demos-g-hgg/. △ Less

Submitted 27 July, 2020; originally announced July 2020.

A satellite orbit drift in binary near-Earth asteroids (66391) 1999 KW4 and (88710) 2001 SL9 -- Indication of the BYORP effect

Authors: P. Scheirich, P. Pravec, P. Kušnirák, K. Hornoch, J. McMahon, D. J. Scheeres, D. Čapek, D. P. Pray, H. Kučáková, A. Galád, J. Vraštil, Yu. N. Krugly, N. Moskovitz, L. D. Avner, B. Skiff, R. S. McMillan, J. A. Larsen, M. J. Brucker, A. F. Tubbiolo, W. R. Cooney, J. Gross, D. Terrell, O. Burkhonov, K. E. Ergashev, Sh. A. Ehgamberdiev , et al. (12 additional authors not shown)

Abstract: We obtained thorough photometric observations of two binary near-Earth asteroids (66391) Moshup = 1999 KW4 and (88710) 2001 SL9 taken from 2000 to 2019 and derived physical and dynamical properties of the binary systems. We found that the data for 1999 KW4 are inconsistent with a constant orbital period and we obtained unique solution with a quadratic drift of the mean anomaly of the satellite of… ▽ More We obtained thorough photometric observations of two binary near-Earth asteroids (66391) Moshup = 1999 KW4 and (88710) 2001 SL9 taken from 2000 to 2019 and derived physical and dynamical properties of the binary systems. We found that the data for 1999 KW4 are inconsistent with a constant orbital period and we obtained unique solution with a quadratic drift of the mean anomaly of the satellite of -0.65 +/- 0.16 deg/yr2 (all quoted uncertainties are 3sigma). This means that the semimajor axis of the mutual orbit of the components of this binary system increases in time with a mean rate of 1.2 +/- 0.3 cm/yr. The data for 2001 SL9 are also inconsistent with a constant orbital period and we obtained two solutions for the quadratic drift of the mean anomaly: 2.8 +/- 0.2 and 5.2 +/- 0.2 deg/yr2, implying that the semimajor axis of the mutual orbit of the components decreases in time with a mean rate of -2.8 +/- 0.2 or -5.1 +/- 0.2 cm/yr for the two solutions, respectively. The expanding orbit of 1999 KW4 may be explained by mutual tides interplaying with binary YORP (BYORP) effect (McMahon and Scheeres, 2010). However, a modeling of the BYORP drift using radar-derived shapes of the binary components predicted a much higher value of the orbital drift than the observed one. It suggests that either the radar-derived shape model of the secondary is inadequate for computing the BYORP effect, or the present theory of BYORP overestimates it. It is possible that the BYORP coefficient has instead an opposite sign than predicted; in that case, the system may be moving into an equilibrium between the BYORP and the tides. In the case of 2001 SL9, the BYORP effect is the only known physical mechanism that can cause the inward drift of its mutual orbit. △ Less

Submitted 4 January, 2021; v1 submitted 13 December, 2019; originally announced December 2019.

Comments: 44 pages, 17 figures. Accepted for publication in Icarus journal

Implicit 3D Orientation Learning for 6D Object Detection from RGB Images

Authors: Martin Sundermeyer, Zoltan-Csaba Marton, Maximilian Durner, Manuel Brucker, Rudolph Triebel

Abstract: We propose a real-time RGB-based pipeline for object detection and 6D pose estimation. Our novel 3D orientation estimation is based on a variant of the Denoising Autoencoder that is trained on simulated views of a 3D model using Domain Randomization. This so-called Augmented Autoencoder has several advantages over existing methods: It does not require real, pose-annotated training data, generalize… ▽ More We propose a real-time RGB-based pipeline for object detection and 6D pose estimation. Our novel 3D orientation estimation is based on a variant of the Denoising Autoencoder that is trained on simulated views of a 3D model using Domain Randomization. This so-called Augmented Autoencoder has several advantages over existing methods: It does not require real, pose-annotated training data, generalizes to various test sensors and inherently handles object and view symmetries. Instead of learning an explicit map** from input images to object poses, it provides an implicit representation of object orientations defined by samples in a latent space. Our pipeline achieves state-of-the-art performance on the T-LESS dataset both in the RGB and RGB-D domain. We also evaluate on the LineMOD dataset where we can compete with other synthetically trained approaches. We further increase performance by correcting 3D orientation estimates to account for perspective errors when the object deviates from the image center and show extended results. △ Less

Submitted 17 July, 2019; v1 submitted 4 February, 2019; originally announced February 2019.

Comments: Code available at: https://github.com/DLR-RM/AugmentedAutoencoder

Kuiper Belt Occultation Predictions

Authors: Wesley C. Fraser, Stephen Gwyn, Chad Trujillo, Andrew W. Stephens, JJ Kavelaars, Michael E. Brown, Federica B. Bianco, Richard P. Boyle, Melissa J. Brucker, Nathan Hetherington, Michael Joner, William C. Keel, Phil P. Langill, Tim Lister, Russet J. McMillan, Leslie Young

Abstract: Here we present observations of 7 large Kuiper Belt Objects. From these observations, we extract a point source catalog with $\sim0.01"$ precision, and astrometry of our target Kuiper Belt Objects with $0.04-0.08"$ precision within that catalog. We have developed a new technique to predict the future occurrence of stellar occultations by Kuiper Belt Objects. The technique makes use of a maximum li… ▽ More Here we present observations of 7 large Kuiper Belt Objects. From these observations, we extract a point source catalog with $\sim0.01"$ precision, and astrometry of our target Kuiper Belt Objects with $0.04-0.08"$ precision within that catalog. We have developed a new technique to predict the future occurrence of stellar occultations by Kuiper Belt Objects. The technique makes use of a maximum likelihood approach which determines the best-fit adjustment to cataloged orbital elements of an object. Using simulations of a theoretical object, we discuss the merits and weaknesses of this technique compared to the commonly adopted ephemeris offset approach. We demonstrate that both methods suffer from separate weaknesses, and thus, together provide a fair assessment of the true uncertainty in a particular prediction. We present occultation predictions made by both methods for the 7 tracked objects, with dates as late as 2015. Finally, we discuss observations of three separate close passages of Quaoar to field stars, which reveal the accuracy of the element adjustment approach, and which also demonstrate the necessity of considering the uncertainty in stellar position when assessing potential occultations. △ Less

Submitted 27 June, 2013; originally announced June 2013.

Comments: 24 pages, 11 figures, accepted to PASP

High Albedos of Low Inclination Classical Kuiper Belt Objects

Authors: M. J. Brucker, W. M. Grundy, J. A. Stansberry, J. R. Spencer, S. S. Sheppard, E. I. Chiang, M. W. Buie

Abstract: We present observations of thermal emission from fifteen transneptunian objects (TNOs) made using the Spitzer Space Telescope. Thirteen of the targets are members of the Classical population: six dynamically hot Classicals, five dynamically cold Classicals, and two dynamically cold inner Classical Kuiper Belt Objects (KBOs). We fit our observations using thermal models to determine the sizes and… ▽ More We present observations of thermal emission from fifteen transneptunian objects (TNOs) made using the Spitzer Space Telescope. Thirteen of the targets are members of the Classical population: six dynamically hot Classicals, five dynamically cold Classicals, and two dynamically cold inner Classical Kuiper Belt Objects (KBOs). We fit our observations using thermal models to determine the sizes and albedos of our targets finding that the cold Classical TNOs have distinctly higher visual albedos than the hot Classicals and other TNO dynamical classes. The cold Classicals are known to be distinct from other TNOs in terms of their color distribution, size distribution, and binarity fraction. The Classical objects in our sample all have red colors yet they show a diversity of albedos which suggests that there is not a simple relationship between albedo and color. As a consequence of high albedos, the mass estimate of the cold Classical Kuiper Belt is reduced from approximately 0.01 Earth masses to approximately 0.001 Earth masses. Our results also increase significantly the sample of small Classical KBOs with known albedos and sizes from 21 to 32 such objects. △ Less

Submitted 1 January, 2009; v1 submitted 22 December, 2008; originally announced December 2008.

Comments: In press, Icarus 26 pages, 4 figures, 6 tables Replaced Fig. 4 with correct axis labels