This paper analyzes anti-phase parametric excitation for a resonant MEMS mirror by independently driving two out-of-plane electrostatic comb-drive actuators positioned at the left and right sides of the MEMS mirror, enabling a fast and reliable start-up from zero amplitude. Both the angular derivative of the comb drives’ capacitance and the square wave driving signals are approximated by complex Fourier series, leading to a nonlinear model that describes the slow evolution of the amplitude and the phase of the MEMS mirror. The proposed model is validated through measurements, demonstrating strong agreement with the analytical results. A detailed discussion on injected and dissipated energy provides an intuitive understanding of the response curve for in-phase and anti-phase excitation signals with various duty cycles. Additionally, the initial start-up behavior of conventional in-phase parametrically excited MEMS mirrors is analyzed and compared to that of MEMS mirrors operated with anti-phase excitation, revealing an improvement of the start-up time by a factor between 8 to 50, depending on the operating point and condition.

The importance of Microelectromechanical Systems (MEMS) is increasing as they cover a variety of applications and therefore represent an efficient solution to various challenges. In different areas, such as Light Detection and Ranging (LiDAR), autonomous driving, medical devices and pico-projectors, they offer a convenient alternative, due to their low cost, low energy consumption, and compact design. This research concentrates on the analysis and enhancement of an existing start-up process for two-dimensional (2D) MEMS mirrors. Influencing parameters are identified in order to speed-up the start-up process and make it more robust and reliable. With the help of mathematical modeling, physical correlations and experimental validation the operation of the system is optimized, where the start-up procedure is tracked in detail and displayed visually. Moreover, the influence of environmental factors, such as temperature variations, on the start-up are analyzed. The results showed that the parameter range could be narrowed down by using mathematical models and therefore it was possible to define a parameter space. The robustness and start-up time was improved by using an optimized parameter set, where a mean total startup time of under 1.5 seconds and a success-rate of 100% was achieved. The performance of the start-up process was also enhanced under temperature variations and vibrations.

The reliability of microelectromechanical systems (MEMS) becomes increasingly important due to the fast market growth, especially in medical, automotive, and aerospace applications, where high reliability performance under harsh environmental conditions is crucial. The aim of this work is to test the behavior and performance of two-dimensional (2D) resonant MEMS mirrors, which are driven electrostatically, under different temperatures. Therefore, six mirrors and their driver application specific integrated circuits (ASIC) are tested inside a temperature chamber, where the temperature is swept between -40°C and 90°C multiple times and different mirror parameters are logged for both axes of the mirrors and the results are evaluated afterwards. Especially the driving frequencies of the axes are monitored. The results show that there is a nearly linear inverse relation between driving frequency of the mirror's fast axis and temperature, as the frequency rises with a decreasing temperature, while for the slow axis the relation between frequency and temperature is more complex. Moreover, the operation of the mirror is stable over the tested temperature range, as no failures are observed.

The reliability assessment and failure analysis of Microelectromechanical System (MEMS) mirrors is a rather new research area, where especially the influence of environmental and mechanical loads on the performance needs to be evaluated. The aim of this work is to test the reliability of two-dimensional (2D) resonant MEMS mirrors, which are driven electrostatically, under applied stress in order to find out after which time and at which stress levels the function of the mirror is affected or a breakage occurs. With the help of the used test setup several mirrors could be tested simultaneously and different mirror parameters could be logged. Measurements have been executed with a duration of approximately one month, where the mirrors were stressed by increasing the oscillation angle over time. The results showed that there are mirror parameters, which may indicate a failure of the mirror before it occurs. Especially a negative drift of the mirror frequency was observed before breakage. Additionally the results showed a correlation between temperature and mirror frequency and information about the distribution of the stress levels at which the mirrors break was gathered.

Synchronized operation of multiple micro-electromechanical systems (MEMS) scanning axes is crucial for applications such as Lissajous scanning. This paper presents a precise linear model and a synchronization control with a fixed frequency ratio for two electrostatically actuated MEMS mirrors driven in parametric resonance. The precise linearized model of the nonlinear resonant mirror is extended for the two-axis synchronization. Based on a master-slave architecture, each mirror is controlled by an individual independent phase-locked loop (PLL) with the displacement current self-sensing, while the duty cycles of the square wave driving signals are either used for amplitude control or phase synchronization to keep a fixed frequency ratio between both mirrors. The dynamics of the controlled nonlinear MEMS mirror at the nominal operation point are analyzed based on a period-to-period energy conservation method, leading to a simple linear model for control design. The derived model and the synchronized operation of the proposed system are verified by measurements, demonstrating an RMS center pixel synchronization error of 0.09 mrad for the master and 0.13 mrad for the slave and providing good maintenance of the high-resolution scanning pattern under environmental influences.

We consider industrial federated learning, a collaboration between a small number of powerful, potentially competing industrial players, mediated by a third party aspiring to improve the service it provides to its customers. We argue that this configuration harbours covert privacy risks that do not arise in e.g. cross-device settings. Companies are very protective of their intellectual property and production processes. Information about changes to their production and the timing of which is to be kept private. We study a scenario in which one of the collaborators infers changes to their competitors' production by detecting potentially subtle temporal data distribution shifts. In this framing, a data distribution shift is always problematic, even if it has no negative effect on training convergence. Thus, our goal is to find means that allow the detection of distributional shifts better than customary evaluation metrics. Based on the assumption that even minor shifts translate into the collaboratively learned machine learning model, the attacker tracks the shared models' internal state with a selection of metrics from literature in order to pick up on relevant changes. In an empirical study on benchmark datasets, we show an honest-but-curious attacker to be capable of detecting subtle distributional shifts on other clients, in some cases long before they become obvious in evaluation.

This paper proposes the direct phase correction phase locked loop (DPCPLL) for simple and robust synchronization of two resonant MEMS mirrors for a Lissajous scan. The DPCPLL, used in a master-slave synchronization structure, runs the slave mirror at the reference frequency and uses the phase of the driving signal to compensate for the synchronization error directly. The DPCPLL merges synchronization control and a PLL, allowing a simple SISO control. The performance of the proposed DPCPLL is assessed based on the vibration immunity of the synchronization. A PID-based DPCPLL and an LQG-based DPCPLL achieve RMS synchronization errors of 87 ns and 69 ns, respectively, which are improvements compared to 118 ns achieved by the conventional structure with a synchronization controller on top of a PLL-operated mirror. The increase of the stability of the Lissajous pattern under a vibration has benefits in applications such as scanning time of flight lidars or Doppler wind lidars.

In this work, we compare multiple end-to-end neural networks that classify and segment numerous anatomies in fetal torso ultrasound (US) images. The novelty of this paper is not restricted by the fact that it extends the scarce literature on the recently proposed nnUNet approach, we are also the first who apply this framework on 2D US data and compare it with various state-of-the-art 2D segmentation models. Our fetal torso dataset comprises two planes – the four chambers of the heart and the three vessel trachea view – with distinct, however, non-mutually exclusive sets of anatomies, which poses another level of complexity. Consequently, besides segmenting observable anatomies, classifying the absence of such anatomies is of crucial importance for researchers and practitioners as well. We find that the nnUNet outperforms numerous state-of-the-art models both in the classification as well as the segmentation task. In more detail, our findings indicate that the nnUNet achieves the highest scores among all evaluation metrics. Finally, we discuss the benefits of the nnUNet and address potential drawbacks of its design regarding 2D segmentation.

This article demonstrates a vibration test for a resonant MEMS scanning system in operation to evaluate the vibration immunity for automotive lidar applications. The MEMS mirror has a reinforcement structure on the backside of the mirror, causing vibration coupling by a mismatch between the center of mass and the rotation axis. An analysis of energy variation is proposed, showing the direction dependency of vibration coupling. Vibration influences are evaluated by transient vibration response and vibration frequency sweep using a single tone vibration for translational y- and z- axis. The measurement results demonstrate standard deviation (STD) amplitude and frequency errors are up to 1.64% and 0.26%, respectively, for 2 $g_\text {rms}$ single tone vibrations on y axis. The simulation results also show a good agreement with both measurements, proving the proposed vibration coupling mechanism of the MEMS mirror. The phased locked loop (PLL) improves the STD amplitude and frequency errors to 0.91% and 0.15% for y axis vibration, corresponding to 44.4% and 43.0% reduction, respectively, showing the benefit of a controlled MEMS mirror for reliable automotive MEMS lidars.

Machine learning has proven to be an enormous asset in industrial settings time and time again. While these methods are responsible for some of the most impressive technical advancements in recent years, machine learning and in particular deep learning, still heavily rely on big datasets, containing all the necessary information to learn a particular task. However, the procurement of useful data often imposes costly adjustments in production in case of internal collection, or has copyright implications in case of external collection. In some cases, the collection fails due to insufficient data quality, or simply availability. Moreover, privacy can be an ethical as well as a legal concern. A promising approach that deals with all of these challenges is to artificially generate data. Unlike real-world data, purely synthetic data does not prompt privacy considerations, allows for better quality control, and in many cases the number of synthetic datapoints is theoretically unlimited. In this work, we explore the utility of synthetic data in industrial settings by outlining several use-cases in the field of Automatic Number Plate Recognition. In all cases synthetic data has the potential of improving the results of the respective deep learning algorithms, substantially reducing the time and effort of data acquisition and preprocessing, and eliminating privacy concerns in a field as sensitive as Automatic Number Plate Recognition.

We describe the verification of a long-range one-dimensional (1D) scanning micro-electro-mechanical systems (MEMS) lidar specifically considering the robustness against external vibration influences. The 1D scanning MEMS lidar exploits a multichannel horizontal line laser to scan the scene vertically for a 10 deg x11 deg horizontal and vertical field of view at a frame rate of up to 29 Hz. To evaluate the robustness against vibrations, a vibration evaluation setup is developed to apply a wideband vibration based on the automotive standard LV124. The vibration tests are performed in three conditions open loop without control and two phase-locked loops (PLLs) with default and high gain settings. The test results demonstrate that vibration can cause wobbly distortion along the scan angle in the open loop case and the PLLs can suppress effectively this influence in the mean and standard deviation of the standard point to surface error up to 69.3% and 90.0%, respectively. This verifies the benefits of the MEMS mirror control, ensuring stable point cloud measurements under vibrations in harsh automotive environments. (C) The Authors. Published by SPIE under a Creative Commons Attribution 4.0 International License.

Federated machine learning frameworks, which take into account confidentiality of distributed data sources are of increasing interest in smart manufacturing. However, the scope of applicability of most such frameworks is restricted in industrial settings due to limitations in the assumptions on the data sources involved. In this work, first, we shed light on the nature of this arising gap between current federated learning and requirements in industrial settings. Our discussion aims at clarifying related notions in emerging sub-disciplines of machine learning, which are partially overlapping. Second, we envision a new confidentiality-preserving approach for smart manufacturing applications based on the more general setting of transfer learning, and envision its implementation in a module-based platform.

The main challenges are discussed together with the lessons learned from past and ongoing research along the development cycle of machine learning systems. This will be done by taking into account intrinsic conditions of nowadays deep learning models, data and software quality issues and human-centered artificial intelligence (AI) postulates, including confidentiality and ethical aspects. The analysis outlines a fundamental theory-practice gap which superimposes the challenges of AI system engineering at the level of data quality assurance, model building, software engineering and deployment. The aim of this paper is to pinpoint research topics to explore approaches to address these challenges.

This contribution presents charge-based capacitive self-sensing with a continuous full state observer for a parametrically driven resonant electrostatic 1D MEMS mirror considering precise and seamless estimation. Based on current integrators, series capacitances or a capacitance network the direct charge self-sensing principles are investigated and compared considering leakage currents, precision and a minimum implementation effort. In comparison to the other charge sensing methods, the proposed methods directly measure the charge changes while the drive voltage is switched on. Since resonant MEMS mirrors are driven by a rectangular signal, the direct self-sensing implies a lack of data when the drive voltage is switched off. A nonlinear observer is also proposed to estimate the full mirror state continuously based on an identified MEMS mirror model. The capacitive charge self-sensing methods achieve overall a high sensing precision of less than 0.14 % RMSE and the observer estimation error of the full state is below 1 % peak-to-peak error regardless of the availability of the charge self-sensing measurements, demonstrates accurate continuous full state estimation. [2021-0130]

This paper proposes a novel self-sensing control concept for resonant MEMS mirrors solely based on the comb-drive current generated by the mirror movement and simple circuitry. Phase errors are immediately compensated by asynchronous switching of the driving voltage using the precise zero crossing detection by the steep current gradient. The mirror amplitude is detected based on the time difference between a comparator threshold crossing of the current signal and the zero crossing of the mirror, while it is controlled by the duty cycle of the driving voltage signal. The proper threshold setting is analyzed regarding the obtained sensitivity and uncertainty of the amplitude detection and is verified by measurements. It is found that even for symmetric out-of-plane comb-drives the scanning direction can be determined utilizing the mode coupling phenomenon of a lightweight MEMS mirror design with reinforcement structure. Experiments show that the proposed control concept results in a low optical pointing uncertainty of 0.52mdeg, which allows 10000 pixels with a precision of 10 sigma at a scanning frequency of 2kHz. Thus a lightweight and simple design of a high performance MEMS mirror is precisely controlled in its oscillation without any additional sensors or complex circuitry.

This paper proposes a phase modulation method for Lissajous scanning systems, which provides adaptive scan pattern design without changing the frame rate or the field of view. Based on a rigorous analysis of Lissajous scanning, phase modulation constrains and a method for pixel calculation are derived. An accurate and simple metric for resolution calculation is proposed based on the area spanned by neighboring pixels and used for scan pattern optimization also considering the scanner dynamics. The methods are implemented using MEMS mirrors for verification of the adaptive pattern shaping, where a 5-fold resolution improvement in a defined region of interest is demonstrated.

This article presents a novel method to derive and identify an accurate small perturbation model of a comb-actuated resonant microelectromechanical system (MEMS) mirror with highly nonlinear dynamics. Besides the nonlinear stiffness and damping, the comb-drives add nonlinearities due to their electrostatic nature and their effect on the dynamic mirror amplitude over frequency behavior. The proposed model is based on a period to period energy conservation and applies for most nonlinearities present in an oscillator such as MEMS mirrors. It is shown that for specific nominal operation points with square wave excitation, the small perturbation model is linear for a wide range. The full dynamics of the derived linear model are parametrized by three constants, that can be estimated by the phase locked loop (PLL), performing a proposed identification method only based on phase measurements. An analysis of control laws usually applied in a PLL provides important information for the proper design of controllers to meet the desired behavior for individual applications.

Accurate phase detection and control of nonlinear resonant micro-opto-electro-mechanical system (MOEMS) mirrors are crucial to achieve stable scanning motions and high resolution imaging as needed in precision applications. This paper proposes a precise phase detection method for an electrostatic actuated MOEMS mirror and a novel digital phase locked loop (PLL) that uses an asynchronous logic for high precision driving and immediate phase compensation, while the clock speed is kept low. The phase of the mirror is detected by an amplified current signal, generated by the movement of the comb drive electrodes, transimpedance amplifiers and a simple comparator circuit. An analysis of the proposed detection method reveals that the pointing uncertainty scales with the product of the driving voltage, the curvature of the comb drive capacitance and the angular velocity of the MOEMS mirror at the zero crossing. The developed fast start-up procedure brings the MOEMS mirror to its maximum amplitude within less than 100 ms with a minimum on required prior knowledge of the used device. The low optical pointing uncertainty of 0.3 mdeg obtained in closed loop operation, allows 19000 pixels with a precision of 10 sigma at a scanning frequency of 2 kHz.

This paper proposes capacitive charge-based self-sensing by integration of the comb drive intrinsic displacement current for resonant electrostatic MEMS mirrors in order to solve the problem of robust feedback for laser scanning in mobile light detection and ranging (Lidar) application. A two-channel switched current integrator circuit is implemented to determine the deflection angle and to distinguish the rotation direction from the asymmetric comb drive charge. Parameters of the MEMS mirror are calibrated with the deflection angle by an optical PSD setup. The resonant electrostatic MEMS mirror is parametrically driven by a square wave high voltage signal, which means, that the charge measurement is only available during the time with non-zero drive signal. From the partly available charge measurements, a nonlinear observer is developed to estimate the mirror state at all time for a potential feedback control. The feasibility for online position estimation is proven by simulation using experimental charge and deflection angle measurements resulting in less than 2% error at full amplitude operation. Finally, the performance of the proposed method is discussed for realization of active MEMS mirror feedback control, overcoming imprecise motions due to structural nonlinearities as well as external disturbances like vibration and climate variation.