Search

We report new computational and experimental evidence of an efficient and astrochemically relevant formation route to formaldehyde (H$_2$CO). This simplest carbonylic compound is central to the formation of complex organics in cold interstellar clouds, and is generally regarded to be formed by the hydrogenation of solid-state carbon monoxide. We demonstrate H$_2$CO formation via the reaction of carbon atoms with amorphous solid water. Crucial to our proposed mechanism is a concerted proton transfer catalyzed by the water hydrogen bonding network. Consequently, the reactions $^3$C + H$_2$O -> $^3$HCOH and $^1$HCOH -> $^1$H$_2$CO can take place with low or without barriers, contrary to the high-barrier traditional internal hydrogen migration. These low barriers or absence thereof explain the very small kinetic isotope effect in our experiments when comparing the formation of H$_2$CO to D$_2$CO. Our results reconcile the disagreement found in the literature on the reaction route: C + H$_2$O -> H$_2$CO., Comment: Accepted for publication in JPCL

Microstructure reconstruction is an important cornerstone to the inverse materials design concept. In this work, a general algorithm is developed to reconstruct a three-dimensional microstructure from given descriptors. Based on two-dimensional (2D) micrographs, this reconstruction algorithm allows valuable insight through spatial visualization of the microstructure and in silico studies of structure-property linkages. The formulation ensures computational efficiency by casting microstructure reconstruction as a gradient-based optimization problem. Herein, the descriptors can be chosen freely, such as spatial correlations or Gram matrices, as long as they are differentiable with respect to the microstructure. Because real microstructure samples are commonly available as 2D microscopy images only, the desired descriptors for the reconstruction process are prescribed on orthogonal 2D slices. This adds a source of noise, which is handled in a new, superior and interpretable manner. The efficiency and applicability of this formulation is demonstrated by various numerical experiments., Comment: pre-print submitted to Acta Materialia

The introduction of machine learning (ML) components in software projects has created the need for software engineers to collaborate with data scientists and other specialists. While collaboration can always be challenging, ML introduces additional challenges with its exploratory model development process, additional skills and knowledge needed, difficulties testing ML systems, need for continuous evolution and monitoring, and non-traditional quality requirements such as fairness and explainability. Through interviews with 45 practitioners from 28 organizations, we identified key collaboration challenges that teams face when building and deploying ML systems into production. We report on common collaboration points in the development of production ML systems for requirements, data, and integration, as well as corresponding team patterns and challenges. We find that most of these challenges center around communication, documentation, engineering, and process and collect recommendations to address these challenges., Comment: 22 pages, 10 figures, 5 tables

For a knot diagram $K$, the classical knot group $\pi_1(K)$ is a free group modulo relations determined by Wirtinger-type relations on the classical crossings. The classical knot group is invariant under the Reidemeister moves. In this paper, we define a set of quotient groups associated to a knot diagram $K$. These quotient groups are invariant under the Reidemeister moves and the set includes the extended knot groups defined by Boden et al and Silver and Williams., Comment: 12 pages, 13 figures

Intelligent navigation among social crowds is an essential aspect of mobile robotics for applications such as delivery, health care, or assistance. Deep Reinforcement Learning emerged as an alternative planning method to conservative approaches and promises more efficient and flexible navigation. However, in highly dynamic environments employing different kinds of obstacle classes, safe navigation still presents a grand challenge. In this paper, we propose a semantic Deep-reinforcement-learning-based navigation approach that teaches object-specific safety rules by considering high-level obstacle information. In particular, the agent learns object-specific behavior by contemplating the specific danger zones to enhance safety around vulnerable object classes. We tested the approach against a benchmark obstacle avoidance approach and found an increase in safety. Furthermore, we demonstrate that the agent could learn to navigate more safely by keeping an individual safety distance dependent on the semantic information., Comment: 7 pages, 5 figures, 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Navigation of mobile robots within crowded environments is an essential task in various use cases, such as delivery, health care, or logistics. Deep Reinforcement Learning (DRL) emerged as an alternative method to replace overly conservative approaches and promises more efficient and flexible navigation. However, Deep Reinforcement Learning is limited to local navigation due to its myopic nature. Previous research works proposed various ways to combine Deep Reinforcement Learning with conventional methods but a common problem is the complexity of highly dynamic environments due to the unpredictability of humans and other objects within the environment. In this paper, we propose a hierarchical waypoint generator, which considers moving obstacles and thus generates safer and more robust waypoints for Deep-Reinforcement-Learning-based local planners. Therefore, we utilize Delaunay Triangulation to encode obstacles and incorporate an extended hybrid A-Star approach to efficiently search for an optimal solution in the time-state space. We compared our waypoint generator against two baseline approaches and outperform them in terms of safety, efficiency, and robustness., Comment: 7 pages, 5 figures

Autonomous navigation of mobile robots is an essential aspect in use cases such as delivery, assistance or logistics. Although traditional planning methods are well integrated into existing navigation systems, they struggle in highly dynamic environments. On the other hand, Deep-Reinforcement-Learning-based methods show superior performance in dynamic obstacle avoidance but are not suitable for long-range navigation and struggle with local minima. In this paper, we propose a Deep-Reinforcement-Learning-based control switch, which has the ability to select between different planning paradigms based solely on sensor data observations. Therefore, we develop an interface to efficiently operate multiple model-based, as well as learning-based local planners and integrate a variety of state-of-the-art planners to be selected by the control switch. Subsequently, we evaluate our approach against each planner individually and found improvements in navigation performance especially for highly dynamic scenarios. Our planner was able to prefer learning-based approaches in situations with a high number of obstacles while relying on the traditional model-based planners in long corridors or empty spaces., Comment: 7 pages, 7 figures

Artificial neural networks (NNs) are one of the most frequently used machine learning approaches to construct interatomic potentials and enable efficient large-scale atomistic simulations with almost ab initio accuracy. However, the simultaneous training of NNs on energies and forces, which are a prerequisite for, e.g., molecular dynamics simulations, can be demanding. In this work, we present an improved NN architecture based on the previous GM-NN model [V. Zaverkin and J. K\"astner, J. Chem. Theory Comput. 16, 5410-5421 (2020)], which shows an improved prediction accuracy and considerably reduced training times. Moreover, we extend the applicability of Gaussian moment-based interatomic potentials to periodic systems and demonstrate the overall excellent transferability and robustness of the respective models. The fast training by the improved methodology is a pre-requisite for training-heavy workflows such as active learning or learning-on-the-fly., Comment: Manuscript accepted for publication in J. Chem. Theory Comput.; Code published at https://gitlab.com/zaverkin_v/gmnn

Machine learning techniques allow a direct mapping of atomic positions and nuclear charges to the potential energy surface with almost ab-initio accuracy and the computational efficiency of empirical potentials. In this work we propose a machine learning method for constructing high-dimensional potential energy surfaces based on feed-forward neural networks. As input to the neural network we propose an extendable invariant local molecular descriptor constructed from geometric moments. Their formulation via pairwise distance vectors and tensor contractions allows a very efficient implementation on graphical processing units (GPUs). The atomic species is encoded in the molecular descriptor, which allows the restriction to one neural network for the training of all atomic species in the data set. We demonstrate that the accuracy of the developed approach in representing both chemical and configurational spaces is comparable to the one of several established machine learning models. Due to its high accuracy and efficiency, the proposed machine-learned potentials can be used for any further tasks, for example the optimization of molecular geometries, the calculation of rate constants or molecular dynamics.

We demonstrate a new image processing methodology for resolving gas bubbles travelling through liquid metal from dynamic neutron radiography images with intrinsically low signal-to-noise ratio. Image pre-processing, denoising and bubble segmentation are described in detail, with practical recommendations. Experimental validation is presented - stationary and moving reference bodies with neutron-transparent cavities are radiographed with imaging conditions similar to the cases with bubbles in liquid metal. The new methods are applied to our experimental data from previous and recent imaging campaigns, and the performance of the methods proposed in this paper is compared against our previously developed methods. Significant improvements are observed as well as the capacity to reliably extract physically meaningful information from measurements performed under highly adverse imaging conditions. The showcased image processing solution and separate elements thereof are readily extendable beyond the present application, and have been made open-source.

The erythrocyte (or red blood cell) sedimentation rate (ESR) is commonly interpreted as a measure of cell aggregation and as a biomarker of inflammation. It is well known that an increase of fibrinogen concentration, an aggregation-inducing protein for erythrocytes, leads to an increase of the sedimentation rate of erythrocytes, which is generally explained through the formation and faster settling of large disjoint aggregates. However, many aspects of erythrocyte sedimentation conform well with the collapse of a colloidal gel rather than with the sedimentation of disjoint aggregates. Using experiments and cell-level numerical simulations, we systematically investigate the dependence of ESR on fibrinogen concentration and its relation to the microstructure of the gel-like erythrocyte suspension. We show that for physiological aggregation interactions, an increase in the attraction strength between cells results in a cell network with larger void spaces. This geometrical change in the network structure occurs due to anisotropic shape and deformability of erythrocytes and leads to an increased gel permeability and faster sedimentation. Our results provide a comprehensive relation between the ESR and the cell-level structure of erythrocyte suspensions and support the gel hypothesis in the interpretation of blood sedimentation.

The erythrocyte sedimentation rate is one of the oldest medical diagnostic methods whose physical mechanisms remain debatable up to date. Using both light microscopy and mesoscale cell-level simulations, we show that erythrocytes form a soft-colloid gel. Furthermore, the high volume fraction of erythrocytes, their deformability, and weak attraction lead to unusual properties of this gel. A theoretical model for the gravitational collapse is developed, whose predictions are in agreement with detailed macroscopic measurements of the interface velocity.

This paper addresses the numerical simulation of self-piercing rivetting processes to join fibre reinforced polymers and sheet metals. Special emphasis is placed on the modelling of the deformation and failure behaviour of the composite material. Different from the simulation of rivetting processes in metals, which requires the modelling of large plastic deformations, the mechanical response of composites is typically governed by intra- and interlaminar damage phenomena. Depending on the polymeric matrix, viscoelastic effects can interfere particularly with the long-term behaviour of the joint. We propose a systematic approach to the modelling of composite laminates, discuss limitations of the used model, and present details of parameter identification. Homogenisation techniques are applied to predict the mechanical behaviour of the composite in terms of effective anisotropic elastic and viscoelastic material properties. In combination with a continuum damage approach this model represents the deformation and failure behaviour of individual laminae. Cohesive zone elements enable the modelling of delamination processes. The parameters of the latter models are identified from experiments. The defined material model for the composite is eventually utilised in the simulation of an exemplary self-piercing rivetting process., Comment: Accepted manuscript

Multi-material lightweight designs, e.g. the combination of aluminum with fiber-reinforced composites, are a key feature for the development of innovative and resource-efficient products. The connection properties of such bi-material interfaces are influenced by the geometric structure on different length scales. In this article a modeling strategy is presented to study the failure behavior of rough interfaces within a computational homogenization scheme. We study different local phenomena and their effects on the overall interface characteristics, e.g. the surface roughness and different local failure types as cohesive failure of the bulk material and adhesive failure of the local interface. Since there is a large separation in the length scales of the surface roughness, which is in the micrometer range, and conventional structural components, we employ a numerical homogenization approach to extract effective traction-separation laws to derive effective interface parameters. Adhesive interface failure is modeled by cohesive elements based on a traction-separation law and cohesive failure of the bulk material is described by an elastic-plastic model with progressive damage evolution., Comment: Accepted manuscript

The application of hybrid composites in lightweight engineering enables the combination of material-specific advantages of fiber-reinforced polymers and classical metals. The interface between the connected materials is of particular interest since failure often initializes in the bonding zone. In this contribution the connection of an aluminum component and a glass fiber-reinforced epoxy is considered on the microscale. The constitutive modeling accounts for adhesive failure of the local interfaces and cohesive failure of the bulk material. Interface failure is represented by cohesive zone models, while the behavior of the polymer is described by an elastic-plastic damage model. A gradient-enhanced formulation is applied to avoid the well-known mesh dependency of local continuum damage models. The application of numerical homogenization schemes allows for the prediction of effective traction-separation relations. Therefore, the influence of the local interface strength and geometry of random rough interfaces on the macroscopical properties is investigated in a numerical study. There is a positive effect of an increased roughness on the effective joint behavior., Comment: Accepted manuscript

Recently, requirements for the explainability of software systems have gained prominence. One of the primary motivators for such requirements is that explainability is expected to facilitate stakeholders' trust in a system. Although this seems intuitively appealing, recent psychological studies indicate that explanations do not necessarily facilitate trust. Thus, explainability requirements might not be suitable for promoting trust. One way to accommodate this finding is, we suggest, to focus on trustworthiness instead of trust. While these two may come apart, we ideally want both: a trustworthy system and the stakeholder's trust. In this paper, we argue that even though trustworthiness does not automatically lead to trust, there are several reasons to engineer primarily for trustworthiness -- and that a system's explainability can crucially contribute to its trustworthiness., Comment: Accepted at the First International Workshop on Requirements Engineering for Explainable Systems (RE4ES) co-located with the 29th IEEE International Requirements Engineering Conference (RE'21)

The prediction of system responses for a given fatigue test bench drive signal is a challenging task, for which linear frequency response function models are commonly used. To account for non-linear phenomena, a novel hybrid model is suggested, which augments existing approaches using Long Short-Term Memory networks. Additional virtual sensing applications of this method are demonstrated. The approach is tested using non-linear experimental data from a servo-hydraulic test rig and this dataset is made publicly available. A variety of metrics in time and frequency domains, as well as fatigue strength under variable amplitudes, are employed in the evaluation., Comment: 20 pages, 11 figures

The recent wave of detections of interstellar aromatic molecules has sparked interest in the chemical behavior of aromatic molecules under astrophysical conditions. In most cases, these detections have been made through chemically related molecules, called proxies, that implicitly indicate the presence of a parent molecule. In this study, we present the results of the theoretical evaluation of the hydrogenation reactions of different aromatic molecules (benzene, pyridine, pyrrole, furan, thiophene, silabenzene, and phosphorine). The viability of these reactions allows us to evaluate the resilience of these molecules to the most important reducing agent in the interstellar medium, the hydrogen atom (H). All significant reactions are exothermic and most of them present activation barriers, which are, in several cases, overcome by quantum tunneling. Instanton reaction rate constants are provided between 50 K and 500 K. For the most efficiently formed radicals, a second hydrogenation step has been studied. We propose that hydrogenated derivatives of furan, pyrrole, and specially 2,3-dihydropyrrole, 2,5-dihydropyrrole, 2,3-dihydrofuran, and 2,5-dihydrofuran are promising candidates for future interstellar detections.

The lack of specifications is a key difference between traditional software engineering and machine learning. We discuss how it drastically impacts how we think about divide-and-conquer approaches to system design, and how it impacts reuse, testing and debugging activities. Traditionally, specifications provide a cornerstone for compositional reasoning and for the divide-and-conquer strategy of how we build large and complex systems from components, but those are hard to come by for machine-learned components. While the lack of specification seems like a fundamental new problem at first sight, in fact software engineers routinely deal with iffy specifications in practice: we face weak specifications, wrong specifications, and unanticipated interactions among components and their specifications. Machine learning may push us further, but the problems are not fundamentally new. Rethinking machine-learning model composition from the perspective of the feature interaction problem, we may even teach us a thing or two on how to move forward, including the importance of integration testing, of requirements engineering, and of design.

This paper presents deep unfolding neural networks to handle inverse problems in photothermal radiometry enabling super resolution (SR) imaging. Photothermal imaging is a well-known technique in active thermography for nondestructive inspection of defects in materials such as metals or composites. A grand challenge of active thermography is to overcome the spatial resolution limitation imposed by heat diffusion in order to accurately resolve each defect. The photothermal SR approach enables to extract high-frequency spatial components based on the deconvolution with the thermal point spread function. However, stable deconvolution can only be achieved by using the sparse structure of defect patterns, which often requires tedious, hand-crafted tuning of hyperparameters and results in computationally intensive algorithms. On this account, Photothermal-SR-Net is proposed in this paper, which performs deconvolution by deep unfolding considering the underlying physics. This enables to super resolve 2D thermal images for nondestructive testing with a substantially improved convergence rate. Since defects appear sparsely in materials, Photothermal-SR-Net applies trained block-sparsity thresholding to the acquired thermal images in each convolutional layer. The performance of the proposed approach is evaluated and discussed using various deep unfolding and thresholding approaches applied to 2D thermal images. Subsequently, studies are conducted on how to increase the reconstruction quality and the computational performance of Photothermal-SR-Net is evaluated. Thereby, it was found that the computing time for creating high-resolution images could be significantly reduced without decreasing the reconstruction quality by using pixel binning as a preprocessing step., Comment: 10 pages, 6 figures

Deep Reinforcement Learning has emerged as an efficient dynamic obstacle avoidance method in highly dynamic environments. It has the potential to replace overly conservative or inefficient navigation approaches. However, the integration of Deep Reinforcement Learning into existing navigation systems is still an open frontier due to the myopic nature of Deep-Reinforcement-Learning-based navigation, which hinders its widespread integration into current navigation systems. In this paper, we propose the concept of an intermediate planner to interconnect novel Deep-Reinforcement-Learning-based obstacle avoidance with conventional global planning methods using waypoint generation. Therefore, we integrate different waypoint generators into existing navigation systems and compare the joint system against traditional ones. We found an increased performance in terms of safety, efficiency and path smoothness especially in highly dynamic environments., Comment: 8 pages, 21 figures

The imagination of the surrounding environment based on experience and semantic cognition has great potential to extend the limited observations and provide more information for mapping, collision avoidance, and path planning. This paper provides a training-based algorithm for mobile robots to perform spatial imagination based on semantic cognition and evaluates the proposed method for the mapping task. We utilize a photo-realistic simulation environment, Habitat, for training and evaluation. The trained model is composed of Resent-18 as encoder and Unet as the backbone. We demonstrate that the algorithm can perform imagination for unseen parts of the object universally, by recalling the images and experience and compare our approach with traditional semantic mapping methods. It is found that our approach will improve the efficiency and accuracy of semantic mapping., Comment: 7 pages, 14 figures

Recently, mobile robots have become important tools in various industries, especially in logistics. Deep reinforcement learning emerged as an alternative planning method to replace overly conservative approaches and promises more efficient and flexible navigation. However, deep reinforcement learning approaches are not suitable for long-range navigation due to their proneness to local minima and lack of long term memory, which hinders its widespread integration into industrial applications of mobile robotics. In this paper, we propose a navigation system incorporating deep-reinforcement-learning-based local planners into conventional navigation stacks for long-range navigation. Therefore, a framework for training and testing the deep reinforcement learning algorithms along with classic approaches is presented. We evaluated our deep-reinforcement-learning-enhanced navigation system against various conventional planners and found that our system outperforms them in terms of safety, efficiency and robustness., Comment: 8 pages, 21 figures, 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)