Refine
Document Type
- Conference Proceeding (213) (remove)
Conference Type
- Konferenzartikel (183)
- Konferenz-Abstract (21)
- Sonstiges (5)
- Konferenz-Poster (2)
- Konferenzband (2)
Keywords
- RoboCup (12)
- Machine Learning (9)
- Deep Leaning (7)
- Heart rhythm model (5)
- Modeling and simulation (5)
- Robustness (4)
- machine learning (4)
- Generative Adversarial Network (3)
- Radar (3)
- cryptography (3)
Institute
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (213) (remove)
Open Access
- Closed Access (82)
- Open Access (75)
- Closed (55)
- Bronze (27)
- Diamond (9)
- Grün (3)
- Gold (1)
Solar energy plays a central role in the energy transition. Clouds generate locally large fluctuations in the generation output of photovoltaic systems, which is a major problem for energy systems such as microgrids, among others. For an optimal design of a power system, this work analyzed the variability using a spatially distributed sensor network at Stuttgart Airport. It has been shown that the spatial distribution partially reduces the variability of solar radiation. A tool was also developed to estimate the output power of photovoltaic systems using irradiation time series and assumptions about the photovoltaic sites. For days with high fluctuations of the estimated photovoltaic power, different energy system scenarios were investigated. It was found the approach can be used to have a more realistic representation of aggregated PV power taking spatial smoothing into account and that the resulting PV power generation profiles provide a good basis for energy system design considerations like battery sizing.
The number of use cases for autonomous vehicles is increasing day by day especially in commercial applications. One important application of autonomous vehicles can be found within the parcel delivery section. Here, autonomous cars can massively help to reduce delivery efforts and time by supporting the courier actively. One important component of course is the autonomous vehicle itself. Nevertheless, beside the autonomous vehicle, a flexible and secure communication architecture also is a crucial key component impacting the overall performance of such system since it is required to allow continuous interactions between the vehicle and the other components of the system. The communication system must provide a reliable and secure architecture that is still flexible enough to remain practical and to address several use cases. In this paper, a robust communication architecture for such autonomous fleet-based systems is proposed. The architecture provides a reliable communication between different system entities while keeping those communications secure. The architecture uses different technologies such as Bluetooth Low Energy (BLE), cellular networks and Low Power Wide Area Network (LPWAN) to achieve its goals.
The desire to connect more and more devices and to make them more intelligent and more reliable, is driving the needs for the Internet of Things more than ever. Such IoT edge systems require sound security measures against cyber-attacks, since they are interconnected, spatially distributed, and operational for an extended period of time. One of the most important requirements for the security in many industrial IoT applications is the authentication of the devices. In this paper, we present a mutual authentication protocol based on Physical Unclonable Functions, where challenge-response pairs are used for both device and server authentication. Moreover, a session key can be derived by the protocol in order to secure the communication channel. We show that our protocol is secure against machine learning, replay, man-in-the-middle, cloning, and physical attacks. Moreover, it is shown that the protocol benefits from a smaller computational, communication, storage, and hardware overhead, compared to similar works.
In recent years, Physical Unclonable Functions (PUFs) have gained significant attraction in the Internet of Things (IoT) for security applications such as cryptographic key generation and entity authentication. PUFs extract the uncontrollable production characteristics of physical devices to generate unique fingerprints for security applications. One common approach for designing PUFs is exploiting the intrinsic features of sensors and actuators such as MEMS elements, which typically exist in IoT devices. This work presents the Cantilever-PUF, a PUF based on a specific MEMS device – Aluminum Nitride (AlN) piezoelectric cantilever. We show the variations of electrical parameters of AlN cantilevers such as resonance frequency, electrical conductivity, and quality factor, as a result of uncontrollable manufacturing process variations. These variations, along with high thermal and chemical stability, and compatibility with silicon technology, makes AlN cantilever a decent candidate for PUF design. We present a cantilever design, which magnifies the effect of manufacturing process variations on electrical parameters. In order to verify our findings, the simulation results of the Monte Carlo method are provided. The results verify the eligibility of AlN cantilever to be used as a basic PUF device for security applications. We present an architecture, in which the designed Cantilever-PUF is used as a security anchor for PUF-enabled device authentication as well as communication encryption.
To deal with frequent power outages in developing countries, people turn to solutions like uninterruptible power supply (UPS), which stores electric energy during normal operating hours and use it to meet energy needs during rolling blackout intervals. Locally produced UPSs of poorer power quality are widely accessible in the marketplaces, and they have a negative impact on power quality. The charging and discharging of the batteries in these UPSs generate significant amount of power losses in weak grid environments. The Smart-UPS is our proposed smart energy metering (SEM) solution for low voltage consumers that is provided by the distribution company. It does not require batteries, therefore there is no power loss or harmonic distortion due to corresponding charging and discharging. Through load flow and harmonic analysis of both traditional UPS and Smart-UPS systems on ETAP, this paper examines their impact on the harmonics and stability of the distribution grid. The simulation results demonstrate that Smart-UPS can assist fixing power quality issues in a developing country like Pakistan by providing cleaner energy than the battery-operated traditional UPSs.
Due to its potential in improving the efficiency of energy supply, smart energy metering (SEM) has become an area of interest with the surge in Internet of Things (IoT). SEM entails remote monitoring and control of the sensors and actuators associated with the energy supply system. This provides a flexible platform to conceive and implement new data driven Demand Side Management (DSM) mechanisms. The IoT enablement allows the data to be gathered and analyzed at requisite granularity. In addition to efficient use of energy resources and provisioning of power, developing countries face an additional challenge of temporal mismatch in generation capacity and load factors. This leads to widespread deployment of inefficient and expensive Uninterruptible Power Supply (UPS) solutions for limited power provisioning during resulting blackouts. Our proposed “Soft-UPS” allows dynamic matching of load and generation through a combination of managed curtailment. This eliminates inefficiencies in the energy and power value chain and allows a data-driven approach to solving a widespread problem in developing countries, simultaneously reducing both upfront and running costs of conventional UPS and storage. A scalable and modular platform is proposed and implemented in this paper. The architecture employs “WiMODino” using LoRaWAN with a “Lite Gateway” and SQLite repository for data storage. Role based access to the system through an android application has also been demonstrated for monitoring and control.
With the surge in global data consumption with proliferation of Internet of Things (IoT), remote monitoring and control is increasingly becoming popular with a wide range of applications from emergency response in remote regions to monitoring of environmental parameters. Mesh networks are being employed to alleviate a number of issues associated with single-hop communication such as low area coverage, reliability, range and high energy consumption. Low-power Wireless Personal Area Networks (LoWPANs) are being used to help realize and permeate the applicability of IoT. In this paper, we present the design and test of IEEE 802.15.4-compliant smart IoT nodes with multi-hop routing. We first discuss the features of the software stack and design choices in hardware that resulted in high RF output power and then present field test results of different baseline network topologies in both rural and urban settings to demonstrate the deployability and scalability of our solution.
Modeling of Random Variations in a Switched Capacitor Circuit based Physically Unclonable Function
(2020)
The Internet of Things (IoT) is expanding to a wide range of fields such as home automation, agriculture, environmental monitoring, industrial applications, and many more. Securing tens of billions of interconnected devices in the near future will be one of the biggest challenges. IoT devices are often constrained in terms of computational performance, area, and power, which demand lightweight security solutions. In this context, hardware-intrinsic security, particularly physically unclonable functions (PUFs), can provide lightweight identification and authentication for such devices. In this paper, random capacitor variations in a switched capacitor PUF circuit are used as a source of entropy to generate unique security keys. Furthermore, a mathematical model based on the ordinary least square method is developed to describe the relationship between random variations in capacitors and the resulting output voltages. The model is used to filter out systematic variations in circuit components to improve the quality of the extracted secrets.
Soiling is an important issue in the renewable energy sector since it can result in significant yield losses, especially in regions with higher pollution or dust levels. To mitigate the impact of soiling on photovoltaic (PV) plants, it is essential to regularly monitor and clean the panels, as well as develop accurate soiling predictions that can affect cleaning strategies and enhance the overall performance of PV power plants. This research focuses on the problem of soiling loss in photovoltaic power plants and the potential to improve the accuracy of soiling predictions. The study examines how soiling can affect the efficiency and productivity of the modules and how to measure and predict soiling using machine learning (ML) algorithms. The research includes analyzing real data from large-scale ground-mounted PV sites and comparing different soiling measurement methods. It was observed that there were some deviations in the real soiling loss values compared to the expected values for some projects in southern Spain, thus, the main goal of this work is to develop machine learning models that could predict the soiling more accurately. The developed models have a low mean square error (MSE), indicating the accuracy and suitability of the models to predict the soiling rates. The study also investigates the impact of different cleaning strategies on the performance of PV power plants and provides a powerful application to predict both the soiling and the number of cleaning cycles.
Elektronische Türschilder zur Darstellung von Informationen sind insbesondere in öffentlichen Gebäuden zwischenzeitlich weit verbreitet. Die Varianz dieser elektronischen Türschilder reicht vom Tablet-basierten Türschild bis hin zum PC-basierten Türschild mit externem Bildschirm. Zumeist werden die Systeme mit 230 V betrieben. Bei einer großen Summe von Türschildern in öffentlichen Gebäuden kann dies zu einem signifikanten Umsatz an Energie führen. Im Rahmen dieses Papers wird die Entwicklung eines energieautarken arbeiten Türschildes vorgestellt, bei dem ein E-Paper-Display zum Einsatz kommt. Das Türschild lässt sich per Smartphone-App und NFC-Schnittstelle konfigurieren. Es wird insbesondere auf das Low-Power-Hardware-Design der Elektronik und energetische Aspekte eingegangen.
Investigation on Bowtie Antennas Operating at Very Low Frequencies for Ground Penetrating Radar
(2023)
The efficiency of Ground Penetrating Radar (GPR) systems significantly depends on the antenna performance as the signal has to propagate through lossy and inhomogeneous media. GPR antennas should have a low operating frequency for greater penetration depth, high gain and efficiency to increase the receiving power and should be compact and lightweight for ease of GPR surveying. In this paper, two different designs of Bowtie antennas operating at very low frequencies are proposed and analyzed.
Many different methods, such as screen printing, gravure, flexography, inkjet etc., have been employed to print electronic devices. Depending on the type and performance of the devices, processing is done at low or high temperature using precursor- or particle-based inks. As a result of the processing details, devices can be fabricated on flexible or non-flexible substrates, depending on their temperature stability. Furthermore, in order to reduce the operating voltage, printed devices rely on high-capacitance electrolytes rather than on dielectrics. The printing resolution and speed are two of the major challenging parameters for printed electronics. High-resolution printing produces small-size printed devices and high-integration densities with minimum materials consumption. However, most printing methods have resolutions between 20 and 50 μm. Printing resolutions close to 1 μm have also been achieved with optimized process conditions and better printing technology.
The final physical dimensions of the devices pose severe limitations on their performance. For example, the channel lengths being of this dimension affect the operating frequency of the thin-film transistors (TFTs), which is inversely proportional to the square of channel length. Consequently, short channels are favorable not only for high-frequency applications but also for high-density integration. The need to reduce this dimension to substantially smaller sizes than those possible with today’s printers can be fulfilled either by developing alternative printing or stamping techniques, or alternative transistor geometries. The development of a polymer pen lithography technique allows scaling up parallel printing of a large number of devices in one step, including the successive printing of different materials. The introduction of an alternative transistor geometry, namely the vertical Field Effect Transistor (vFET), is based on the idea to use the film thickness as the channel length, instead of the lateral dimensions of the printed structure, thus reducing the channel length by orders of magnitude. The improvements in printing technologies and the possibilities offered by nanotechnological approaches can result in unprecedented opportunities for the Internet of Things (IoT) and many other applications. The vision of printing functional materials, and not only colors as in conventional paper printing, is attractive to many researchers and industries because of the added opportunities when using flexible substrates such as polymers and textiles. Additionally, the reduction of costs opens new markets. The range of processing techniques covers laterally-structured and large-area printing technologies, thermal, laser and UV-annealing, as well as bonding techniques, etc. Materials, such as conducting, semiconducting, dielectric and sensing materials, rigid and flexible substrates, protective coating, organic, inorganic and polymeric substances, energy conversion and energy storage materials constitute an enormous challenge in their integration into complex devices.
Team description papers of magmaOffenburg are incremental in the sense that each year we address a different topic of our team and the tools around our team. In this year’s team description paper we focus on the architecture of the software. It is a main factor for being able to keep the code maintainable even after 15 years of development. We also describe how we make sure that the code follows this architecture.
In this paper, we study the runtime performance of symmetric cryptographic algorithms on an embedded ARM Cortex-M4 platform. Symmetric cryptographic algorithms can serve to protect the integrity and optionally, if supported by the algorithm, the confidentiality of data. A broad range of well-established algorithms exists, where the different algorithms typically have different properties and come with different computational complexity. On deeply embedded systems, the overhead imposed by cryptographic operations may be significant. We execute the algorithms AES-GCM, ChaCha20-Poly1305, HMAC-SHA256, KMAC, and SipHash on an STM32 embedded microcontroller and benchmark the execution times of the algorithms as a function of the input lengths.
Cryptographic protection of messages requires frequent updates of the symmetric cipher key used for encryption and decryption, respectively. Protocols of legacy IT security, like TLS, SSH, or MACsec implement rekeying under the assumption that, first, application data exchange is allowed to stall occasionally and, second, dedicated control messages to orchestrate the process can be exchanged. In real-time automation applications, the first is generally prohibitive, while the second may induce problematic traffic patterns on the network. We present a novel seamless rekeying approach, which can be embedded into cyclic application data exchanges. Although, being agnostic to the underlying real-time communication system, we developed a demonstrator emulating the widespread industrial Ethernet system PROFINET IO and successfully use this rekeying mechanism.
Towards a Formal Verification of Seamless Cryptographic Rekeying in Real-Time Communication Systems
(2022)
This paper makes two contributions to the verification of communication protocols by transition systems. Firstly, the paper presents a modeling of a cyclic communication protocol using a synchronized network of transition systems. This protocol enables seamless cryptographic rekeying embedded into cyclic messages. Secondly, we test the protocol using the model checking verification technique.
Current training methods for deep neural networks boil down to very high dimensional and non-convex optimization problems which are usually solved by a wide range of stochastic gradient descent methods. While these approaches tend to work in practice, there are still many gaps in the theoretical understanding of key aspects like convergence and generalization guarantees, which are induced by the properties of the optimization surface (loss landscape). In order to gain deeper insights, a number of recent publications proposed methods to visualize and analyze the otimization surfaces. However, the computational cost of these methods are very high, making it hardly possible to use them on larger networks. In this paper, we present the GradVis Toolbox, an open source library for efficient and scalable visualization and analysis of deep neural network loss landscapes in Tesorflow and PyTorch. Introducing more efficient mathematical formulations and a novel parallelization scheme, GradVis allows to plot 2d and 3d projections of optimization surfaces and trajectories, as well as high resolution second order gradient information for large networks.
In this paper, we propose a unified approach for network pruning and one-shot neural architecture search (NAS) via group sparsity. We first show that group sparsity via the recent Proximal Stochastic Gradient Descent (ProxSGD) algorithm achieves new state-of-the-art results for filter pruning. Then, we extend this approach to operation pruning, directly yielding a gradient-based NAS method based on group sparsity. Compared to existing gradient-based algorithms such as DARTS, the advantages of this new group sparsity approach are threefold. Firstly, instead of a costly bilevel optimization problem, we formulate the NAS problem as a single-level optimization problem, which can be optimally and efficiently solved using ProxSGD with convergence guarantees. Secondly, due to the operation-level sparsity, discretizing the network architecture by pruning less important operations can be safely done without any performance degradation. Thirdly, the proposed approach finds architectures that are both stable and well-performing on a variety of search spaces and datasets.
We demonstrate how to exploit group sparsity in order to bridge the areas of network pruning and neural architecture search (NAS). This results in a new one-shot NAS optimizer that casts the problem as a single-level optimization problem and does not suffer any performance degradation from discretizating the architecture.
The use of artificial intelligence continues to impact a broad variety of domains, application areas, and people. However, interpretability, understandability, responsibility, accountability, and fairness of the algorithms' results - all crucial for increasing humans' trust into the systems - are still largely missing. The purpose of this seminar is to understand how these components factor into the holistic view of trust. Further, this seminar seeks to identify design guidelines and best practices for how to build interactive visualization systems to calibrate trust.
In this paper we report on further success of our work to develop a multi-method energy optimization which works with a digital twin concept. The twin concept serves to replicate production processes of different kinds of production companies, including complex energy systems and test market interactions to then use them for model predictive optimizing. The presented work finally reports about the performed flexibility assessment leading to a flexibility audit with a list of measures and the impact of energy optimizations made related to interactions with the local power grid i.e., the exchange node of the low voltage distribution grid. The analysis and continuous exploration of flexibilities as well as the exchange with energy markets require a “guide” leading to continuous optimization with a further tool like the Flexibility Survey and Control Panel helping decision-making processes on the day-ahead horizon for real production plants or the investment planning to improve machinery, staff schedules and production
infrastructure.
The twin concept is increasingly used for optimization tasks in the context of Industry 4.0 and digitization. The twin concept can also help small and medium-sized enterprises (SME) to exploit their energy flexibility potential and to achieve added value by appropriate energy marketing. At the same time, this use of flexibility helps to realize a climate-neutral energy supply with high shares of renewable energies. The digital twin reflects real production, power flows and market influences as a computer model, which makes it possible to simulate and optimize on-site interventions and interactions with the energy market without disturbing the real production processes. This paper describes the development of a generic model library that maps flexibility-relevant components and processes of SME, thus simplifying the creation of a digital twin. The paper also includes the development of an experimental twin consisting of SME hardware components and a PLC-based SCADA system. The experimental twin provides a laboratory environment in which the digital twin can be tested, further developed and demonstrated on a laboratory scale. Concrete implementations of such a digital twin and experimental twin are described as examples.
The present work describes an extension of current slope estimation for parameter estimation of permanent magnet synchronous machines operated at inverters. The area of operation for current slope estimation in the individual switching states of the inverter is limited due to measurement noise, bandwidth limitation of the current sensors and the commutation processes of the inverter's switching operations. Therefore, a minimum duration of each switching state is necessary, limiting the final area of operation of a robust current slope estimation. This paper presents an extension of existing current slope estimation algorithms resulting in a greater area of operation and a more robust estimation result.
In this work a method for the estimation of current slopes induced by inverters operating interior permanent magnet synchronous machines is presented. After the derivation of the estimation algorithm, the requirements for a suitable sensor setup in terms of accuracy, dynamic and electromagnetic interference are discussed. The boundary conditions for the estimation algorithm are presented with respect to application within high power traction systems. The estimation algorithm is implemented on a field programmable gateway array. This moving least-square algorithm offers the advantage that it is not dependent on vectors and therefore not every measured value has to be stored. The summation of all measured values leads to a significant reduction of the required storage units and thus decreases the hardware requirements. The algorithm is designed to be calculated within the dead time of the inverter. Appropriate countermeasures for disturbances and hardware restrictions are implemented. The results are discussed afterwards.
The following describes a new method for estimating the parameters of an interior permanent magnet synchronous machine (IPMSM). For the estimation of the parameters the current slopes caused by the switching of the inverter are used to determine the unknowns of the system equations of the electrical machine. The angle and current dependence of the machine parameters are linearized within a PWM cycle. By considering the different switching states of the inverter, several system equations can be derived and a solution can be found within one PWM cycle. The use of test signals and filter-based approaches is avoided. The derived algorithm is explained and validated with measurements on a test bench.
A Novel Approach of High Dynamic Current Control of Interior Permanent Magnet Synchronous Machines
(2019)
Harmonic-afflicted effects of permanent magnet synchronous machines with high power density are hardly faced by traditional current PI controllers, due to limited controller bandwidth. As a consequence, currents and lastly torque ripples appear. In this paper, a new deadbeat current controller architecture has been presented, which is capable to encounter the effects of these harmonics. This new control algorithm, here named “Hybrid-Deadbeat-Controller”, combines the stability and the low steady-state errors offered by common PI regulators with the high dynamic offered by the deadbeat control. Therefore, a novel control algorithm is proposed, capable of either compensating the current harmonics in order to get smoother currents or to control a varying reference value to achieve a smoother torque. The information needed to calculate the optimal reference currents is based on an online parameter estimation feeding an optimization algorithm to achieve an optimal torque output and will be investigated in future research. In order to ensure the stability of the controller over the whole area of operation even under the influence of effects changing the system’s parameter, this work as well focusses on the robustness of the “hybrid” dead beat controller.
In this study, various imaging algorithms for the localization of objects have been investigated. Therefore, an Ultra-Wideband (UWB) radar based experimental setup with a circular antenna array is designed as part of this work. This concept could be particularly useful in microwave medical imaging applications. In order to validate its applicability in microwave imaging, different imaging algorithms have been evaluated and compared by means of our experimental setup. Accurate imaging results have been achieved with our system under multiple test-scenarios.
In this study, an approach to a microwave-based radar system for the localization of objects has been proposed. This could be particularly useful in microwave imaging applications such as cardiac catheter detection. An experimental system is defined and realized with the selection of an appropriate antenna design. Hardware control functions and different imaging algorithms are implemented as well. The functionality of this measurement setup has been analyzed considering multiple testscenarios and it is proved to be capable of locating multiple objects as well as expanded objects.
This paper presents the use of model predictive control (MPC) based approach for peak shaving application of a battery in a Photovoltaic (PV) battery system connected to a rural low voltage gird. The goals of the MPC are to shave the peaks in the PV feed-in and the grid power consumption and at the same time maximize the use of the battery. The benefit to the prosumer is from the maximum use of the self-produced electricity. The benefit to the grid is from the reduced peaks in the PV feed-in and the grid power consumption. This would allow an increase in the PV hosting and the load hosting capacity of the grid.
The paper presents the mathematical formulation of the optimal control problem
along with the cost benefit analysis. The MPC implementation scheme in the
laboratory and experiment results have also been presented. The results show
that the MPC is able to track the deviation in the weather forecast and operate
the battery by solving the optimal control problem to handle this deviation.
Due to the Covid-19 pandemic, the RoboCup WorldCup 2021 was held completely remotely. For this competition the Webots simulator (https://cyberbotics.com/) was used, so all teams needed to transfer their robot to the simulation. This paper describes our experiences during this process as well as a genetic learning approach to improve our walk engine to allow a more stable and faster movement in the simulation. Therefore we used a docker setup to scale easily. The resulting movement was one of the outstanding features that finally led to the championship title.
Sweaty has already participated several times in RoboCup soccer competitions (Adult Size). Now the work is focused on stabilizing the gait. Moreover, we would like to overcome the constraints of a ZMP-algorithm that has a horizontal footplate as precondition for the simplification of the equations. In addition we would like to switch between impedance and position control with a fuzzy-like algorithm that might help to minimize jerks when Sweaty’s feet touch the ground.
Sweaty has already participated four times in RoboCup soccer competitions (Adult Size) and came second three times. While 2016 Sweaty needed a lot of luck to be finalist, 2017 Sweaty was a serious adversary in the preliminary rounds. In 2018 Sweaty showed up in the final with some lack of experience and room for improvements, but not without any chance. This paper describes the intended improvements of the humanoid adult size robot Sweaty in order to qualify for the RoboCup 2019 adult size competition.
Autonomous driving is disrupting the automotive industry as we know it today. For this, fail-operational behavior is essential in the sense, plan, and act stages of the automation chain in order to handle safety-critical situations on its own, which currently is not reached with state-of-the-art approaches.The European ECSEL research project PRYSTINE realizes Fail-operational Urban Surround perceptION (FUSION) based on robust Radar and LiDAR sensor fusion and control functions in order to enable safe automated driving in urban and rural environments. This paper showcases some of the key exploitable results (e.g., novel Radar sensors, innovative embedded control and E/E architectures, pioneering sensor fusion approaches, AI-controlled vehicle demonstrators) achieved until its final year 3.
Generative adversarial networks (GANs) provide state-of-the-art results in image generation. However, despite being so powerful, they still remain very challenging to train. This is in particular caused by their highly non-convex optimization space leading to a number of instabilities. Among them, mode collapse stands out as one of the most daunting ones. This undesirable event occurs when the model can only fit a few modes of the data distribution, while ignoring the majority of them. In this work, we combat mode collapse using second-order gradient information. To do so, we analyse the loss surface through its Hessian eigenvalues, and show that mode collapse is related to the convergence towards sharp minima. In particular, we observe how the eigenvalues of the are directly correlated with the occurrence of mode collapse. Finally, motivated by these findings, we design a new optimization algorithm called nudged-Adam (NuGAN) that uses spectral information to overcome mode collapse, leading to empirically more stable convergence properties.
Transformer models have recently attracted much interest from computer vision researchers and have since been successfully employed for several problems traditionally addressed with convolutional neural networks. At the same time, image synthesis using generative adversarial networks (GANs) has drastically improved over the last few years. The recently proposed TransGAN is the first GAN using only transformer-based architectures and achieves competitive results when compared to convolutional GANs. However, since transformers are data-hungry architectures, TransGAN requires data augmentation, an auxiliary super-resolution task during training, and a masking prior to guide the self-attention mechanism. In this paper, we study the combination of a transformer-based generator and convolutional discriminator and successfully remove the need of the aforementioned required design choices. We evaluate our approach by conducting a benchmark of well-known CNN discriminators, ablate the size of the transformer-based generator, and show that combining both architectural elements into a hybrid model leads to better results. Furthermore, we investigate the frequency spectrum properties of generated images and observe that our model retains the benefits of an attention based generator.
Seismic data processing involves techniques to deal with undesired effects that occur during acquisition and pre-processing. These effects mainly comprise coherent artefacts such as multiples, non-coherent signals such as electrical noise, and loss of signal information at the receivers that leads to incomplete traces. In this work, we employ a generative solution, since it can explicitly model complex data distributions and hence, yield to a better decision-making process. In particular, we introduce diffusion models for multiple removal. To that end, we run experiments on synthetic and on real data, and we compare the deep diffusion performance with standard algorithms. We believe that our pioneer study not only demonstrates the capability of diffusion models, but also opens the door to future research to integrate generative models in seismic workflows.
Generative adversarial networks are the state of the art approach towards learned synthetic image generation. Although early successes were mostly unsupervised, bit by bit, this trend has been superseded by approaches based on labelled data. These supervised methods allow a much finer-grained control of the output image, offering more flexibility and stability. Nevertheless, the main drawback of such models is the necessity of annotated data. In this work, we introduce an novel framework that benefits from two popular learning techniques, adversarial training and representation learning, and takes a step towards unsupervised conditional GANs. In particular, our approach exploits the structure of a latent space (learned by the representation learning) and employs it to condition the generative model. In this way, we break the traditional dependency between condition and label, substituting the latter by unsupervised features coming from the latent space. Finally, we show that this new technique is able to produce samples on demand keeping the quality of its supervised counterpart.
Facial image manipulation is a generation task where the output face is shifted towards an intended target direction in terms of facial attribute and styles. Recent works have achieved great success in various editing techniques such as style transfer and attribute translation. However, current approaches are either focusing on pure style transfer, or on the translation of predefined sets of attributes with restricted interactivity. To address this issue, we propose FacialGAN, a novel framework enabling simultaneous rich style transfers and interactive facial attributes manipulation. While preserving the identity of a source image, we transfer the diverse styles of a target image to the source image. We then incorporate the geometry information of a segmentation mask to provide a fine-grained manipulation of facial attributes. Finally, a multi-objective learning strategy is introduced to optimize the loss of each specific tasks. Experiments on the CelebA-HQ dataset, with CelebAMask-HQ as semantic mask labels, show our model’s capacity in producing visually compelling results in style transfer, attribute manipulation, diversity and face verification. For reproducibility, we provide an interactive open-source tool to perform facial manipulations, and the Pytorch implementation of the model.
A fundamental and still largely unsolved question in the context of Generative Adversarial Networks is whether they are truly able to capture the real data distribution and, consequently, to sample from it. In particular, the multidimensional nature of image distributions leads to a complex evaluation of the diversity of GAN distributions. Existing approaches provide only a partial understanding of this issue, leaving the question unanswered. In this work, we introduce a loop-training scheme for the systematic investigation of observable shifts between the distributions of real training data and GAN generated data. Additionally, we introduce several bounded measures for distribution shifts, which are both easy to compute and to interpret. Overall, the combination of these methods allows an explorative investigation of innate limitations of current GAN algorithms. Our experiments on different data-sets and multiple state-of-the-art GAN architectures show large shifts between input and output distributions, showing that existing theoretical guarantees towards the convergence of output distributions appear not to be holding in practice.
Generative convolutional deep neural networks, e.g. popular GAN architectures, are relying on convolution based up-sampling methods to produce non-scalar outputs like images or video sequences. In this paper, we show that common up-sampling methods, i.e. known as up-convolution or transposed convolution, are causing the inability of such models to reproduce spectral distributions of natural training data correctly. This effect is independent of the underlying architecture and we show that it can be used to easily detect generated data like deepfakes with up to 100% accuracy on public benchmarks. To overcome this drawback of current generative models, we propose to add a novel spectral regularization term to the training optimization objective. We show that this approach not only allows to train spectral consistent GANs that are avoiding high frequency errors. Also, we show that a correct approximation of the frequency spectrum has positive effects on the training stability and output quality of generative networks.
Recent deep learning based approaches have shown remarkable success on object segmentation tasks. However, there is still room for further improvement. Inspired by generative adversarial networks, we present a generic end-to-end adversarial approach, which can be combined with a wide range of existing semantic segmentation networks to improve their segmentation performance. The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss. In addition, we train our generator employing stochastic weight averaging fashion, which further enhances the predicted output label maps leading to state-of-the-art results. We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.
The term “attribute transfer” refers to the tasks of altering images in such a way, that the semantic interpretation of a given input image is shifted towards an intended direction, which is quantified by semantic attributes. Prominent example applications are photo realistic changes of facial features and expressions, like changing the hair color, adding a smile, enlarging the nose or altering the entire context of a scene, like transforming a summer landscape into a winter panorama. Recent advances in attribute transfer are mostly based on generative deep neural networks, using various techniques to manipulate images in the latent space of the generator. In this paper, we present a novel method for the common sub-task of local attribute transfers, where only parts of a face have to be altered in order to achieve semantic changes (e.g. removing a mustache). In contrast to previous methods, where such local changes have been implemented by generating new (global) images, we propose to formulate local attribute transfers as an inpainting problem. Removing and regenerating only parts of images, our “Attribute Transfer Inpainting Generative Adversarial Network” (ATI-GAN) is able to utilize local context information to focus on the attributes while keeping the background unmodified resulting in visually sound results.
In this preliminary report, we present a simple but very effective technique to stabilize the training of CNN based GANs. Motivated by recently published methods using frequency decomposition of convolutions (eg Octave Convolutions), we propose a novel convolution scheme to stabilize the training and reduce the likelihood of a mode collapse. The basic idea of our approach is to split convolutional filters into additive high and low frequency parts, while shifting weight updates from low to high during the training. Intuitively, this method forces GANs to learn low frequency coarse image structures before descending into fine (high frequency) details. Our approach is orthogonal and complementary to existing stabilization methods and can simply plugged into any CNN based GAN architecture. First experiments on the CelebA dataset show the effectiveness of the proposed method.
Background: The application of high-frequency ablation is used for the treatment of tachycardia arrhythmias and is a respected method. Ablation with high frequency current leads to the targeted heat destruction of myocardial tissue at specific sites and thus prevents the pathological propagation of excitation through these structures.
Purpose: The aim of this study was to simulate heat propagation during RF ablation with modeled electrodes in different sizes and materials. The simulation was performed on atrioventricular node re-entry tachycardia (AVNRT), atrioventricular re-entry tachycardia (AVRT) and atrial flutter (AFL).
Methods: Using the modeling and simulation software CST, ablation catheters with 4 mm and 8 mm tip electrodes were modeled from gold and platinum for each. The designed catheters correspond to the manufacturer"s specifications of Medtronic, Biotronik and Osypka. The catheters were integrated into the Offenburg heart rhythm model to simulate and compare the heat propagation during an ablation application, which also takes into account the blood flow in the four heart chambers. A power of 5 W - 40 W was simulated for the 4 mm electrodes and a power of 50 W - 80 W for the 8 mm electrodes.
Results: During the simulated HF ablation application, the temperature at the ablation electrode was measured at different powers. This is 40.67°C at 5 W, 44.34°C at 10 W, 51.76°C at 20 W, 59.0°C at 30 W, and 66.33°C at 40 W. The measured temperature during 40 W application is 39.5°C at 0,5 mm depth in the myocardium and 37.5°C at 2 mm depth.
In the simulation, the 8 mm platinum electrode reached an ablation temperature of 72.85°C at its tip during an applied power of 60 W. In contrast, the 8 mm platinum electrode reached a depth of 5 mm at 39.5 C° and at a depth of 2 mm at 37.5 °C. In contrast, the 8 mm gold electrode reached a temperature of 64.66°C with the same performance. This is due to the thermal properties of gold, which has a better thermal conductivity than platinum.
Conclusions: CST offers the possibility to carry out a static and dynamic simulation of a heart model and the ablation electrodes integrated in it during an HF ablation. In variation with different electrode sizes and materials, therapy methods for the treatment of AVNRT, AVRT and AFL can be optimized
Narrowband IoT (NB-IoT) as a radio access technology for the cellular Internet of Things (cIoT) is getting more traction due to attractive system parameters, new proposals in the 3 rd Generation Partnership Project (3GPP) Release 14 for reduced power consumption and ongoing world-wide deployment. As per 3GPP, the low-power and wide-area use cases in 5G specification will be addressed by the early NB-IoT and Long-Term Evolution for Machines (LTE-M) based technologies. Since these cIoT networks will operate in a spatially distributed environment, there are various challenges to be addressed for tests and measurements of these networks. To meet these requirements, unified emulated and field testbeds for NB-IoT-networks were developed and used for extensive performance measurements. This paper analyses the results of these measurements with regard to RF coverage, signal quality, latency, and protocol consistency.
Wireless communication technologies play a major role to enable megatrends like Internet of Things (IoT) and Industry 4.0. The Narrowband Wireless WAN (NBWWAN) introduced to meet the long range and low power requirements of spatially distributed wireless communication use cases. These networks introduce additional challenges in testing because the network topology and RF characteristics become particularly complex and thus a multitude of different scenarios must be tested. This paper describes the infrastructure for automated testing of radio communication and for systematic measurements of the network performance of NBWWAN.
Spatially Distributed Wireless Networks (SDWN) are one of the basic technologies for the Internet of Things (IoT) and (Industrial) Internet of Things (IIoT) applications. These SDWN for many of these applications has strict requirements such as low cost, simple installation and operations, and high potential flexibility and mobility. Among the different Narrowband Wireless Wide Area Networking (NBWWAN) technologies, which are introduced to address these categories of wireless networking requirements, Narrowband Internet of Things (NB-IoT) is getting more traction due to attractive system parameters, energy-saving mode of operation with low data rates and bandwidth, and its applicability in 5G use cases. Since several technologies are available and because the underlying use cases come with various requirements, it is essential to perform a systematic comparative analysis of competing technologies to choose the right technology. It is also important to perform testing during different phases of the system development life cycle. This paper describes the systematic test environment for automated testing of radio communication and systematic measurements of the performance of NB-IoT.
One of the main requirements of spatially distributed Internet of Things (IoT) solutions is to have networks with wider coverage to connect many low-power devices. Low-Power Wide-Area Networks (LPWAN) and Cellular IoT(cIOT) networks are promising candidates in this space. LPWAN approaches are based on enhanced physical layer (PHY) implementations to achieve long range such as LoRaWAN, SigFox, MIOTY. Narrowband versions of cellular network offer reduced bandwidth and, simplified node and network management mechanisms, such as Narrow Band IoT (NB-IoT) and Long-Term Evolution for Machines (LTE-M). Since the underlying use cases come with various requirements it is essential to perform a comparative analysis of competing technologies. This article provides systematic performance measurement and comparison of LPWAN and NB-IoT technologies in a unified testbed, also discusses the necessity of future fifth generation (5G) LPWAN solutions.
Printed electronics (PE) is a fast growing technology with promising applications in wearables, smart sensors and smart cards since it provides mechanical flexibility, low-cost, on-demand and customizable fabrication. To secure the operation of these applications, True Random Number Generators (TRNGs) are required to generate unpredictable bits for cryptographic functions and padding. However, since the additive fabrication process of PE circuits results in high intrinsic variation due to the random dispersion of the printed inks on the substrate, constructing a printed TRNG is challenging. In this paper, we exploit the additive customizable fabrication feature of inkjet printing to design a TRNG based on electrolyte-gated field effect transistors (EGFETs). The proposed memory-based TRNG circuit can operate at low voltages (≤ 1 V ), it is hence suitable for low-power applications. We also propose a flow which tunes the printed resistors of the TRNG circuit to mitigate the overall process variation of the TRNG so that the generated bits are mostly based on the random noise in the circuit, providing a true random behaviour. The results show that the overall process variation of the TRNGs is mitigated by 110 times, and the simulated TRNGs pass the National Institute of Standards and Technology Statistical Test Suite.
Electrode modelling and simulation of diagnostic and pulmonary vein isolation in atrial fibrillation
(2022)
Seismic data has often missing traces due to technical acquisition or economical constraints. A compete dataset is crucial in several processing and inversion techniques. Deep learning algorithms, based on convolutional neural networks (CNNs), have shown alternative solutions that overcome limitation of traditional interpolation methods e.g. data regularity, linearity assumption, etc. There are two different paradigms of CNN methods for seismic interpolation. The first one, so-called deep prior interpolation (DPI), trains a CNN to map random noise to a complete seismic image using only the decimated image itself. The second one, referred as standard deep learning method, trains a CNN to map a decimated seismic image into a complete one using a dataset of complete and artificially decimated images. Within this research, we systematically compare the performance of both methods for different quantities of regular and irregular missing traces using 4 datasets. We evaluate the results of both methods using 5 well-known metrics. We found that DPI method performs better than the standard method if the percentage of missing traces is low (10%) and otherwise if the level of decimation is high (50%).
In this work, we explore three deep learning algorithms apply to seismic interpolation: deep prior image (DPI), standard, and generative adversarial networks (GAN). The standard and GAN approaches rely on a dataset of complete and decimated seismic images for the training process, while the DPI method learns from a decimated image itself, without training images. We carry out two main experiments, considering 10%, 30%, and 50% of regular and irregular decimation. The first tests the optimal situation for the GAN and the standard approaches, where training and testing images are from the same dataset. The second tests the ability of GAN and standard methods to learn simultaneously from three datasets, and generalize to a fourth dataset not used during training. The standard method provides the best results in the first experiment, when the training distribution is similar to the testing one. In this situation, the DPI approach reports the second best results. In the second experiment, the standard method shows the ability to learn simultaneously and effectively three data distributions for the regular case. In the irregular case, the DPI approach is more effective. The GAN approach is the less effective of the three deep learning methods in both experiments.
Seismic data processing relies on multiples attenuation to improve inversion and interpretation. Radon-based algorithms are often used for multiples and primaries discrimination. Deep learning, based on convolutional neural networks (CNNs), has shown encouraging applications for demultiple that could mitigate Radon-based challenges. In this work, we investigate new strategies to train a CNN for multiples removal based on different loss functions. We propose combined primaries and multiples labels in the loss for training a CNN to predict primaries, multiples, or both simultaneously. Moreover, we investigate two distinctive training methods for all the strategies: UNet based on minimum absolute error (L1) training, and adversarial training (GAN-UNet). We test the trained models with the different strategies and methods on 400 synthetic data. We found that training to predict multiples, including the primaries …
An important step in seismic data processing to improve inversion and interpretation is multiples attenuation. Radon-based algorithms are often used for discriminating primaries and multiples. Recently, deep learning (DL), based on convolutional neural networks (CNNs) has shown promising results in demultiple that could mitigate the challenges of Radon-based methods. In this work, we investigate new different strategies to train a CNN for multiples removal based on different loss functions. We propose combined primaries and multiples labels in the loss for training a CNN to predict primaries, multiples, or both simultaneously. We evaluate the performance of the CNNs trained with the different strategies on 400 clean and noisy synthetic data, considering 3 metrics. We found that training a CNN to predict the multiples and then subtracting them from the input image is the most effective strategy for demultiple. Furthermore, including the primaries labels as a constraint during the training of multiples prediction improves the results. Finally, we test the strategies on a field dataset. The CNNs trained with different strategies report competitive results on real data compared with Radon demultiple. As a result, effectively trained CNN models can potentially replace Radon-based demultiple in existing workflows.
It is common practice to apply padding prior to convolution operations to preserve the resolution of feature-maps in Convolutional Neural Networks (CNN). While many alternatives exist, this is often achieved by adding a border of zeros around the inputs. In this work, we show that adversarial attacks often result in perturbation anomalies at the image boundaries, which are the areas where padding is used. Consequently, we aim to provide an analysis of the interplay between padding and adversarial attacks and seek an answer to the question of how different padding modes (or their absence) affect adversarial robustness in various scenarios.
Currently, many theoretical as well as practically relevant questions towards the transferability and robustness of Convolutional Neural Networks (CNNs) remain unsolved. While ongoing research efforts are engaging these problems from various angles, in most computer vision related cases these approaches can be generalized to investigations of the effects of distribution shifts in image data. In this context, we propose to study the shifts in the learned weights of trained CNN models. Here we focus on the properties of the distributions of dominantly used 3×3 convolution filter kernels. We collected and publicly provide a dataset with over 1.4 billion filters from hundreds of trained CNNs, using a wide range of datasets, architectures, and vision tasks. In a first use case of the proposed dataset, we can show highly relevant properties of many publicly available pre-trained models for practical applications: I) We analyze distribution shifts (or the lack thereof) between trained filters along different axes of meta-parameters, like visual category of the dataset, task, architecture, or layer depth. Based on these results, we conclude that model pre-training can succeed on arbitrary datasets if they meet size and variance conditions. II) We show that many pre-trained models contain degenerated filters which make them less robust and less suitable for fine-tuning on target applications. Data & Project website: https://github.com/paulgavrikov/cnn-filter-db.
Deep learning models are intrinsically sensitive to distribution shifts in the input data. In particular, small, barely perceivable perturbations to the input data can force models to make wrong predictions with high confidence. An common defense mechanism is regularization through adversarial training which injects worst-case perturbations back into training to strengthen the decision boundaries, and to reduce overfitting. In this context, we perform an investigation of 3 × 3 convolution filters that form in adversarially- trained models. Filters are extracted from 71 public models of the ℓ ∞ -RobustBench CIFAR-10/100 and ImageNet1k leaderboard and compared to filters extracted from models built on the same architectures but trained without robust regularization. We observe that adversarially-robust models appear to form more diverse, less sparse, and more orthogonal convolution filters than their normal counterparts. The largest differences between robust and normal models are found in the deepest layers, and the very first convolution layer, which consistently and predominantly forms filters that can partially eliminate perturbations, irrespective of the architecture.
Recent work has investigated the distributions of learned convolution filters through a large-scale study containing hundreds of heterogeneous image models. Surprisingly, on average, the distributions only show minor drifts in comparisons of various studied dimensions including the learned task, image domain, or dataset. However, among the studied image domains, medical imaging models appeared to show significant outliers through "spikey" distributions, and, therefore, learn clusters of highly specific filters different from other domains. Following this observation, we study the collected medical imaging models in more detail. We show that instead of fundamental differences, the outliers are due to specific processing in some architectures. Quite the contrary, for standardized architectures, we find that models trained on medical data do not significantly differ in their filter distributions from similar architectures trained on data from other domains. Our conclusions reinforce previous hypotheses stating that pre-training of imaging models can be done with any kind of diverse image data.
An Empirical Investigation of Model-to-Model Distribution Shifts in Trained Convolutional Filters
(2021)
We present first empirical results from our ongoing investigation of distribution shifts in image data used for various computer vision tasks. Instead of analyzing the original training and test data, we propose to study shifts in the learned weights of trained models. In this work, we focus on the properties of the distributions of dominantly used 3x3 convolution filter kernels. We collected and publicly provide a data set with over half a billion filters from hundreds of trained CNNs, using a wide range of data sets, architectures, and vision tasks. Our analysis shows interesting distribution shifts (or the lack thereof) between trained filters along different axes of meta-parameters, like data type, task, architecture, or layer depth. We argue, that the observed properties are a valuable source for further investigation into a better understanding of the impact of shifts in the input data to the generalization abilities of CNN models and novel methods for more robust transfer-learning in this domain.
Neural networks have a number of shortcomings. Amongst the severest ones is the sensitivity to distribution shifts which allows models to be easily fooled into wrong predictions by small perturbations to inputs that are often imperceivable to humans and do not have to carry semantic meaning. Adversarial training poses a partial solution to address this issue by training models on worst-case perturbations. Yet, recent work has also pointed out that the reasoning in neural networks is different from humans. Humans identify objects by shape, while neural nets mainly employ texture cues. Exemplarily, a model trained on photographs will likely fail to generalize to datasets containing sketches. Interestingly, it was also shown that adversarial training seems to favorably increase the shift toward shape bias. In this work, we revisit this observation and provide an extensive analysis of this effect on various architectures, the common L_2-and L_-training, and Transformer-based models. Further, we provide a possible explanation for this phenomenon from a frequency perspective.
The EREMI project is a 2-year project funded under the ERASMUS+ framework programme and its team has developed and will validate an advanced higher education program, including life-long learning, on the interdisciplinary topic of resource efficiency in manufacturing industries and the overall system optimization of low or not digitized physical infrastructure. All of these will be achieved by applying IoT technologies towards efficient industrial systems, and by utilizing a high-level educated human capital on these economically, politically, and technically crucial and highly relevant topics for the rapidly developing industries and economies of intensively economically and industrially transforming countries - Bulgaria, North Macedonia, and Romania. Efficiency will be attained by utilizing the experience and expertise of the involved German partner organisation.
The importance of machine learning has been increasing dramatically for years. From assistance systems to production optimisation to support the health sector, almost every area of daily life and industry comes into contact with machine learning. Besides all the benefits that ML brings, the lack of transparency and the difficulty in creating traceability pose major risks. While there are solutions that make the training of machine learning models more transparent, traceability is still a major challenge. Ensuring the identity of a model is another challenge. Unnoticed modification of a model is also a danger when using ML. One solution is to create an ML birth certificate and an ML family tree secured by blockchain technology. Important information about training and changes to the model through retraining can be stored in a blockchain and accessed by any user to create more security and traceability about an ML model.
Despite the success of convolutional neural networks (CNNs) in many academic benchmarks for computer vision tasks, their application in the real-world is still facing fundamental challenges. One of these open problems is the inherent lack of robustness, unveiled by the striking effectiveness of adversarial attacks. Current attack methods are able to manipulate the network's prediction by adding specific but small amounts of noise to the input. In turn, adversarial training (AT) aims to achieve robustness against such attacks and ideally a better model generalization ability by including adversarial samples in the trainingset. However, an in-depth analysis of the resulting robust models beyond adversarial robustness is still pending. In this paper, we empirically analyze a variety of adversarially trained models that achieve high robust accuracies when facing state-of-the-art attacks and we show that AT has an interesting side-effect: it leads to models that are significantly less overconfident with their decisions, even on clean data than non-robust models. Further, our analysis of robust models shows that not only AT but also the model's building blocks (like activation functions and pooling) have a strong influence on the models' prediction confidences. Data & Project website: https://github.com/GeJulia/robustness_confidences_evaluation
Despite the success of convolutional neural networks (CNNs) in many academic benchmarks for computer vision tasks, their application in the real-world is still facing fundamental challenges. One of these open problems is the inherent lack of robustness, unveiled by the striking effectiveness of adversarial attacks. Adversarial training (AT) is often considered as a remedy to train more robust networks. In this paper, we empirically analyze a variety of adversarially trained models that achieve high robust accuracies when facing state-of-the-art attacks and we show that AT has an interesting side-effect: it leads to models that are significantly less overconfident with their decisions even on clean data than non-robust models. Further, our analysis of robust models shows that not only AT but also the model's building blocks (like activation functions and pooling) have a strong influence on the models' prediction confidences.
Over the last years, Convolutional Neural Networks (CNNs) have been the dominating neural architecture in a wide range of computer vision tasks. From an image and signal processing point of view, this success might be a bit surprising as the inherent spatial pyramid design of most CNNs is apparently violating basic signal processing laws, i.e. Sampling Theorem in their down-sampling operations. However, since poor sampling appeared not to affect model accuracy, this issue has been broadly neglected until model robustness started to receive more attention. Recent work in the context of adversarial attacks and distribution shifts, showed after all, that there is a strong correlation between the vulnerability of CNNs and aliasing artifacts induced by poor down-sampling operations. This paper builds on these findings and introduces an aliasing free down-sampling operation which can easily be plugged into any CNN architecture: FrequencyLowCut pooling. Our experiments show, that in combination with simple and Fast Gradient Sign Method (FGSM) adversarial training, our hyper-parameter free operator substantially improves model robustness and avoids catastrophic overfitting. Our code is available at https://github.com/GeJulia/flc_pooling
Many commonly well-performing convolutional neural network models have shown to be susceptible to input data perturbations, indicating a low model robustness. Adversarial attacks are thereby specifically optimized to reveal model weaknesses, by generating small, barely perceivable image perturbations that flip the model prediction. Robustness against attacks can be gained for example by using adversarial examples during training, which effectively reduces the measurable model attackability. In contrast, research on analyzing the source of a model’s vulnerability is scarce. In this paper, we analyze adversarially trained, robust models in the context of a specifically suspicious network operation, the downsampling layer, and provide evidence that robust models have learned to downsample more accurately and suffer significantly less from aliasing than baseline models.
The automatic processing of handwritten forms remains a challenging task, wherein detection and subsequent classification of handwritten characters are essential steps. We describe a novel approach, in which both steps - detection and classification - are executed in one task through a deep neural network. Therefore, training data is not annotated by hand, but manufactured artificially from the underlying forms and yet existing datasets. It can be demonstrated that this single-task approach is superior in comparison to the state-of-the-art two task approach. The current study focuses on hand-written Latin letters and employs the EMNIST data set. However, limitations were identified with this data set, necessitating further customization. Finally, an overall recognition rate of 88.28% was attained on real data obtained from a written exam.
Background: Transesophageal left atrial (LA) pacing and transesophageal LA ECG recording are semi-invasive techniques for diagnostic and therapy of supraventricular rhythm disturbance. Cardiac resynchronization therapy (CRT) with right atrial (RA) sensed biventricular pacing is an established therapy for heart failure patients with reduced left ventricular (LV) ejection fraction, sinus rhythm and interventricular electrical desynchronization.
Purpose: The aim of the study was to evaluate electromagnetic and voltage pacing fields of the combination of RA pacing, LA pacing and biventricular pacing in patients with long interatrial and interventricular electrical desynchronization.
Methods: The modelling and electromagnetic simulations of transesophageal LA pacing in combination with RA pacing and biventricular pacing would be staged and analyzed with the CST (Computer Simulation Technology) software. Different electrodes were modelled in order to simulate different types of bipolar pacing in the 3D-CAD Offenburg heart rhythm model: The bipolar Solid S (Biotronik) electrode where modelled for RA pacing and right ventricular (RV) pacing, Attain 4194 (Medtronic) for LV pacing and TO8 (Osypka) multipolar esophageal electrode with hemispheric electrodes for LA pacing.
Results: The pacemaker amplitudes for the electromagnetic pacing simulations were performed with 3 V for RA pacing, 1.5 V for RV pacing, 50 V for LA pacing and 3V for LV pacing with pacing impulse duration of 0.5 ms for RA, RV and LV pacing and 10 ms for LA pacing. The atrioventricular pacing delay after RA pacing was 140 ms. The different pacing modes AAI, VVI, DDD, DDD0V and DDD0D were evaluated for the analysis of the electric pacing field propagation of pacemaker, CRT and LA pacing. The pacing results were compared at minimum (LOW) and maximum (HIGH) parameter settings. While the LOW setting produced fewer tetrahedral and more inaccurate results, the HIGH setting produced many tetrahedral and therefore more accurate results.
Conclusions: The simulation of the combination of transesophageal LA pacing with RA sensed biventricular pacing is possible with the Offenburg heart rhythm model. The new temporary 4-chamber pacing method may be additional useful method in CRT non-responders with long interatrial electrical delay.
The Transport Layer Security protocol is a widespread cryptographic protocol designed to provide secure communication over insecure networks by providing authenticity, integrity, and confidentiality. As a first step, in the TLS Handshake Protocol a common master secret is negotiated. In many configurations, this step makes considerable use of asymmetric cryptographic algorithms. It seems to be a prevalent assumption that the use of such asymmetric cryptographic algorithms is unsuitable for resource-constrained devices. Therefore, the work at hand analyzes the runtime performance of the TLS vl.2 session establishments on an embedded ARM Cortex-M4 platform. We measure the execution time to generate and parse session establishment messages for the client and server sides. In particular, we study the impact of different elliptic curves used for the ephemeral Diffie-Hellman key exchange and the impact of different lengths and subject public key algorithms of certification paths. Our analysis shows that the use of asymmetric cryptographic algorithms is well possible on resource-constrained devices, if carefully chosen and well implemented. This allows the use of the well-proven TLS protocol also for applications from the (Industrial) Internet of Things, including Fieldbus communication.
Machine learning (ML) has become highly relevant in applications across all industries, and specialists in the field are sought urgently. As it is a highly interdisciplinary field, requiring knowledge in computer science, statistics and the relevant application domain, experts are hard to find. Large corporations can sweep the job market by offering high salaries, which makes the situation for small and medium enterprises (SME) even worse, as they usually lack the capacities both for attracting specialists and for qualifying their own personnel. In order to meet the enormous demand in ML specialists, universities now teach ML in specifically designed degree programs as well as within established programs in science and engineering. While the teaching almost always uses practical examples, these are somewhat artificial or outdated, as real data from real companies is usually not available. The approach reported in this contribution aims to tackle the above challenges in an integrated course, combining three independent aspects: first, teaching key ML concepts to graduate students from a variety of existing degree programs; second, qualifying working professionals from SME for ML; and third, applying ML to real-world problems faced by those SME. The course was carried out in two trial periods within a government-funded project at a university of applied sciences in south-west Germany. The region is dominated by SME many of which are world leaders in their industries. Participants were students from different graduate programs as well as working professionals from several SME based in the region. The first phase of the course (one semester) consists of the fundamental concepts of ML, such as exploratory data analysis, regression, classification, clustering, and deep learning. In this phase, student participants and working professionals were taught in separate tracks. Students attended regular classes and lab sessions (but were also given access to e-learning materials), whereas the professionals learned exclusively in a flipped classroom scenario: they were given access to e-learning units (video lectures and accompanying quizzes) for preparation, while face-to-face sessions were dominated by lab experiments applying the concepts. Prior to the start of the second phase, participating companies were invited to submit real-world problems that they wanted to solve with the help of ML. The second phase consisted of practical ML projects, each tackling one of the problems and worked on by a mixed team of both students and professionals for the period of one semester. The teams were self-organized in the ways they preferred to work (e.g. remote vs. face-to-face collaboration), but also coached by one of the teaching staff. In several plenary meetings, the teams reported on their status as well as challenges and solutions. In both periods, the course was monitored and extensive surveys were carried out. We report on the findings as well as the lessons learned. For instance, while the program was very well-received, professional participants wished for more detailed coverage of theoretical concepts. A challenge faced by several teams during the second phase was a dropout of student members due to upcoming exams in other subjects.
Despite the success of convolutional neural networks (CNNs) in many computer vision and image analysis tasks, they remain vulnerable against so-called adversarial attacks: Small, crafted perturbations in the input images can lead to false predictions. A possible defense is to detect adversarial examples. In this work, we show how analysis in the Fourier domain of input images and feature maps can be used to distinguish benign test samples from adversarial images. We propose two novel detection methods: Our first method employs the magnitude spectrum of the input images to detect an adversarial attack. This simple and robust classifier can successfully detect adversarial perturbations of three commonly used attack methods. The second method builds upon the first and additionally extracts the phase of Fourier coefficients of feature-maps at different layers of the network. With this extension, we are able to improve adversarial detection rates compared to state-of-the-art detectors on five different attack methods. The code for the methods proposed in the paper is available at github.com/paulaharder/SpectralAdversarialDefense
Aerosol particles play an important role in the climate system by absorbing and scattering radiation and influencing cloud properties. They are also one of the biggest sources of uncertainty for climate modeling. Many climate models do not include aerosols in sufficient detail. In order to achieve higher accuracy, aerosol microphysical properties and processes have to be accounted for. This is done in the ECHAM-HAM global climate aerosol model using the M7 microphysics model, but increased computational costs make it very expensive to run at higher resolutions or for a longer time. We aim to use machine learning to approximate the microphysics model at sufficient accuracy and reduce the computational cost by being fast at inference time. The original M7 model is used to generate data of input-output pairs to train a neural network on it. By using a special logarithmic transform we are able to learn the variables tendencies achieving an average score of . On a GPU we achieve a speed-up of 120 compared to the original model.
Estimation of Scattering and Transfer Parameters in Stratified Dispersive Tissues of the Human Torso
(2021)
The aim of this study is to understand the effect of the various layers of biological tissues on electromagnetic radiation in a certain frequency range. Understanding these effects could prove crucial in the development of dynamic imaging systems under operating environments during catheter ablation in the heart. As the catheter passes through some arterial paths in the region of interest inside the heart through the aorta, a three-dimensional localization of the catheter is required. In this paper, a study is given on the detection of the catheter by using electromagnetic waves. Therefor, an appropriate model for the layers of the human torso is defined and simulated without and with an inserted electrode.
The present work ties in with the problem of bicycle road assessment that is currently done using expensive special measuring vehicles. Our alternative approach for road condition assessment is to mount a sensor device on a bicycle which sends accelerometer and gyroscope data via WiFi to a classification server. There, a prediction model determines road type and condition based on the sensor data. For the classification task, we compare different machine learning methods with each other, whereby validation accuracies of 99% can be achieved with deep residual networks such as InceptionTime. The main contribution of this work with respect to comparable work is that we achieve excellent accuracies on a realistic dataset classifying road conditions into nine distinct classes that are highly relevant for practice.
This study focuses on the autonomous navigation and mapping of indoor environments using a drone equipped only with a monocular camera and height measurement sensors. A visual SLAM algorithm was employed to generate a preliminary map of the environment and to determine the drone's position within the map. A deep neural network was utilized to generate a depth image from the monocular camera's input, which was subsequently transformed into a point cloud to be projected into the map. By aligning the depth point cloud with the map, 3D occupancy grid maps were constructed by using ray tracing techniques to get a precise depiction of obstacles and the surroundings. Due to the absence of IMU data from the low-cost drone for the SLAM algorithm, the created maps are inherently unscaled. However, preliminary tests with relative navigation in unscaled maps have revealed potential accuracy issues, which can only be overcome by incorporating additional information from the given sensors for scale estimation.
Modern industrial production is heavily dependent on efficient workflow processes and automation. The steady flow of raw materials as well as the separation of vital parts and semi-finished products are at the core of these automated procedures. Commonly used systems for this work are bowl feeders, which separate the parts and material by a combination of mechanical vibration and friction. The production of these tools, especially the design of the ramping spiral, is delicate and time-consuming work, as the shape, slope, and material must be carefully adjusted for the corresponding parts. In this work, we propose an automated approach, making use of optimization procedures from artificial intelligence, to design the spiral ramps of the bowl feeders. Therefore, the whole system and considered parts are physically simulated and the optimized geometry is subsequently exported into a CAD system for the actual building, respectively printing. The employment of evolutionary optimization gives the need to develop a mathematical model for the whole setup and find an efficient representation of integral features.
The precise positioning of mobile systems is a prerequisite for any autonomous behavior, in an industrial environment as well as for field robotics. The paper describes the set up for an experimental platform and its use for the evaluation of simultaneous localization and mapping (SLAM) algorithms. Two approaches are compared. First, a local method based on point cloud matching and integration of inertial measurement units is evaluated. Subsequent matching makes it possible to create a three-dimensional point cloud that can be used as a map in subsequent runs. The second approach is a full SLAM algorithm, based on graph relaxation models, incorporating the full sensor suite of odometry, inertial sensors, and 3D laser scan data.
Die Positionierung mobiler Systeme mit hoher Genauigkeit ist eine Voraussetzung für intelligentes autonomes Verhalten, sowohl in der Feldrobotik als auch in industriellen Umgebungen. Dieser Beitrag beschreibt den Aufbau einer Roboterplattform und ihre Verwendung für den Test und die Bewertung von Kalman-Filter-Konfigurationen. Der Aufbau wurde mit einem mobilen Roboter Husky A200 und einem LiDAR-Sensor (Light Detection and Ranging) realisiert. Zur Verifizierung des vorgeschlagenen Aufbaus wurden fünf verschiedene Szenarien ausgearbeitet. Mit denen wurden die Filter auf ihre Leistungsfähigkeit hinsichtlich der Genauigkeit der Positionsbestimmung getestet.
Evaluierung von Kalman Filter Konfigurationen zur Roboterlokaliserung mittels Sensordatenfusion
(2023)
In dieser Arbeit werden drei verschiedene Konfigurationen der von Tom Moore, für das Robot Operating System, entwickelte Kalman-Filter vorgestellt. Diese bilden die Grundlage für eine Lokalisierung mittels Sensorfusion in dem verwendeten ROS-Framework. Ziel dieser Arbeit ist der Aufbau und die Verifikation einer Lokalisierung für ein mobiles Robotersystem Husky A200 der Firma Clearpath Robotics. Hierzu wurden die Möglichkeiten des bestehenden Systems untersucht und mehrere Versionen von Lokalisierungsfiltern konfiguriert. Am an Ende, wird eine Verifikation der Ergebnisse in verschiedenen Szenarien gegeneinandergestellt. Hierzu werden die Ergebnisse einer Variante des Extended Kalman-Filters in 2D (EKF2D), eine Variante des Unscented Kalman-Filter in 2D (UKF2D) und eine Variante des Extended Kalman-Filters in 3D (EKF3D) verifiziert und verglichen. Die Untersuchungen ergaben das der EKF2D die besten und robustesten Ergebnisse für eine Lokalisierung erbringt, trotz, im Vergleich zu der UKF2D Variante, 17,3 % höhere Endpositionsabweichung aufweist. Die in diesem Projekt gewählte EKF3D Konfigurationsvariante eignet sich, wegen seinen starken Ungenauigkeiten in der Höhenbestimmung nicht für eine aussagekräftige Positionsbestimmung.
Mit der Implementierung sowie einer anschließenden aussagekräftigen Evaluierung, soll das, visuelle-inertiale Kartierungs- und Lokalisierungssystem maplab analysiert werden. Hierbei basiert die Kartierung bzw. Lokalisierung auf der Detektion von Umgebungsmerkmalen. Neben der Möglichkeit der Kartenerstellung besteht ferner die Option, mehrere Karten zu fusionieren und somit weitreichende Gebiete zu kartieren sowie für weitere Datenauswertungen zu nutzen. Aufgrund der Durchführung und Bewertung der Ergebnisse in unterschiedlichen Anwendungsszenarien zeigt sich, dass maplab besonders zur Kartierung von Räumen bzw. kleinen Gebäudekomplexen geeignet ist. Die Möglichkeit der Kartenfusionierung bietet weiterhin die Option, den Informationsgehalt von Karten zu erhöhen, welches die Effektivität für eine anschließende Lokalisierung steigert. Bei wachsender Kartierungsgröße hingegen zeigt sich jedoch eine Vergrößerung geometrischer Inkonsistenzen.
A novel approach for synchronization and calibration of a camera and an inertial measurement unit (IMU) in the research-oriented visual-inertial mapping-and localization-framework maplab is presented. Mapping and localization are based on detecting different features in the environment. In addition to the possibility of creating single-case maps, the included algorithms allow merging maps to increase mapping accuracy and obtain large-scale maps. Furthermore, the algorithms can be used to optimize the collected data. The preliminary results show that after appropriate calibration and synchronization maplab can be used efficiently for mapping, especially in rooms and small building environments.
The visual-inertial mapping and localization system maplab is analyzed by its implementation and subsequent evaluation. The mapping or localization is based on environmental feature detection. In addition to creating maps, there is also the option of fusion of several maps and thus mapping extensive areas and using them for further analysis of data. In this way, various software tools can be used to optimize the existing data sets.
Two sensor components are needed: an inertial measuring unit (IMU) and a monochrome camera, which are combined by a hardware rig and put into operation for the analysis of the visual-inertial system. System calibration is crucial for precision and system functioning and is based on nonlinear dynamic state estimation. This ensures the best possible estimate of the position of the environmental feature and the map. Maplab is particularly suitable for mapping rooms or small building complexes as the implementation and evaluation of the results in different application scenarios show. Special emphasis is laid on the evaluation of larger scenarios, in which is shown, that the system is struggling to keep up geometric consistencies and thus provide an accurate map.
Object Detection and Mapping with Unmanned Aerial Vehicles Using Convolutional Neural Networks
(2021)
Significant progress has been made in the field of deep learning through intensive research over the last decade. So-called convolutional neural networks are an essential component of this research. In this type of neural network, the mathematical convolution operator is used to extract characteristics or anomalies. The purpose of this work is to investigate the extent to which it is possible in certain initial settings to input aerial recordings and flight data of Unmanned Aerial Vehicles (UAVs) in the architecture of a neural network and to detect and map an object. Using the calculated contours or dimensions of the so-called bounding boxes, the position of the objects can be determined relative to the current UAV location.
The paper describes a systematic approach for a precise short-time cloud coverage prediction based on an optical system. We present a distinct pre-processing stage that uses a model based clear sky simulation to enhance the cloud segmentation in the images. The images are based on a sky imager system with fish-eye lens optic to cover a maximum area. After a calibration step, the image is rectified to enable linear prediction of cloud movement. In a subsequent step, the clear sky model is estimated on actual high dynamic range images and combined with a threshold based approach to segment clouds from sky. In the final stage, a multi hypothesis linear tracking framework estimates cloud movement, velocity and possible coverage of a given photovoltaic power station. We employ a Kalman filter framework that efficiently operates on the rectified images. The evaluation on real world data suggests high coverage prediction accuracy above 75%.
The applicability of characteristics of local magnetic fields for more precise determination of localization of subjects and/or objects in indoor environments, such as railway stations, airports, exhibition halls, showrooms, or shopping centers, is considered. An investigation has been carried out to find out whether and how low-cost magnetic field sensors and mobile robot platforms can be used to create maps that improve the accuracy and robustness of later navigation with smartphones or other devices.
The aim of this work is the application and evaluation of a method to visually detect markers at a distance of up to five meters and determine their real-world position. Combinations of cameras and lenses with different parameters were studied to determine the optimal configuration. Based on this configuration, camera images were taken after proper calibration. These images are then transformed into a bird's eye view using a homography matrix. The homography matrix is calculated with four-point pairs as well as with coordinate transformations. The obtained images show the ground plane un distorted, making it possible to convert a pixel position into a real-world position with a conversion factor. The proposed approach helps to effectively create data sets for training neural networks for navigation purposes.
Bei dem vorgestellten Ansatz soll der Auftreffpunkt des Pfeils durch die Kreuzkorrelation von Audio-Signalen bestimmt werden. Das Auftreffen des Pfeils erzeugt ein charakteristisches Geräusch, welches von mehreren Mikrofonen in bestimmter Anordnung um die Dartscheibe herum in elektrische Signale umgewandelt wird. Mithilfe der Schallgeschwindigkeit und den Zeitdifferenzen, welche die Schallwelle zu den einzelnen Mikrofonen benötigt soll dann der Auftreffpunkt berechnet werden.
This paper presents an approach for implementing an automated hit detection and score calculation system for a steel dartboard using a standard webcam. First, the rectilinear field separations of the dartboard are described mathematically by means of line slopes and are than stored. These slopes serve as a basis for later score calculation. In addition, thrown darts have to be detected and the pixel at which the dart cuts the dartboard has to be determined. When this information is known, a comparison is made using the line slopes, allowing the field number of the hit to be detected. The decision for single, double or triple hit is made by evaluating the defined colors on the dartboard. All these functions are then packaged in a Matlab GUI.
Sweaty has already participated several times in RoboCup soccer competitions (Adult Size). Now the work is focused coordinating the play of two robots. Moreover, we are working on stabilizing the gait by adding additional sensor information. An ongoing work is the optimization of the control strategy by balancing between impedance and position control. By minimizing the jerk, gait and overall gameplay should improve significantly.