Using independent vector analysis (IVA) to analyze and find subgroups in functional magnetic resonance imaging (fMRI) and functional ultrasound (fUS) data requires a lot of manual labour. Recently methods like subgroup identification using IVA (SI-IVA) and IVA for common subspace identification (IVA-CS) have tried to reduce this labour through automation. However, both methods did not test for accuracy. This thesis shows through simulations that these proposed methods are not accurate or robust enough to be trusted and that spectral clustering is a better alternative for automatic subgroup identification. Spectral clustering is then incorporated into the analysis of experimental fUS data of two groups of mice to try and identify these automatically. In this analysis, adaptive constrained IVA (acIVA) was used to incorporate references, further improving the interpretability of the results as components are directly linked to prior constraints.
However, applying subgroup analysis showed that the mice could not be clustered based on their response to the stimuli. Still, spectral clustering is more accurate in the simulations making it a promising alternative for automatic subgroup identification. Furthermore, combining spectral clustering with acIVA makes the results more interpretable due to constrained components not being subject to permutation ambiguity.

This report explains the processes behind the design, construction and testing of a high-voltage power grid simulator and visualizer. For this, a system based on data found in a paper was created. This paper describes an aggregated grid for the Dutch high-voltage power grid in 2018. The main use case of this system would be for educative purposes, where interested students and potential students can come, play and learn at the same time while making interactive decisions and seeing how their actions are influencing the system as a whole. The subsystem in this report is tasked with the integration of the data from the forecasting & scenario module and the inputs from the hardware module, furthermore it calculates the power flow using a Gauss-Seidel algorithm and visualizes its results alongside the forecasting in Godot

In the framework of this MSc thesis, the analysis and design guidelines for part of the quasi-optical system of a proposed astronomical instrument, called TIFUUN, are presented. The TIFUUN instrument is an imaging spectrometer, planned to be placed in the ASTE telescope to perform ground-based astronomical observations in the mm-submm wavelength regime. Part of the instrument development involves the design of the quasi-optical system coupling the radiation from the telescope’s main dish to the spectrometer array. In this thesis, the focal plane array of antennas, as well as the first component of the quasi-optical chain are analyzed and two different design approaches are presented, as candidate geometries. Each of them satisfies the requirements of different science surveys targeted by TIFUUN. The first examined architecture is comprised of a focal plane array of on-chip feeding elements under a single hyper-hemispherical lens, which is then coupled geometrically to a hyperbolic lens. The second geometry is instead comprised of an array of integrated elliptical lenses, with a single on-chip feeding element per lens, diffractively coupled to a hyperbolic lens. The methodologies to efficiently analyze these kinds of geometries are Geometrical Optics (GO) combined with analysis in reception for the first design approach and Coherent Fourier Optics (CFO) for the second one. During the design process, the performance of both architectures is optimized throughout the field of view, using methodologies to correct for phase aberrations, such as feed displacement inside the lenses and symmetric shaping of dielectric surfaces. The design guidelines provided and insights obtained during this MSc project will be utilized to develop the quasi-optical system of the TIFUUN instrument, within the limited space of its cryostat.

Artificial Intelligence has become a dominant part of our lives, however, complex artificial intelligence models tend to use a lot of energy, computationally complex operations, and a lot of memory resources. Therefore, it excluded a whole class of hardware in its applicability. Namely, relatively resource-constrained low-cost hardware. This paper investigates learning methods that are potentially better suited for these types of devices: the forward-forward algorithm and Hebbian learning rules. The results are compared to backpropagation with equivalent network configurations, training hyperparameters and internal data types on different types of low-cost hardware. Backpropagation has consistently outperformed other algorithms in various tests. It exhibits higher accuracy, faster training, and faster inference compared to forward-forward models. However, forward-forward models can come close to matching backpropagation's performance, but they suffer from longer training times and decreased performance with multi-layer networks. Additionally, a poorly trained forward-forward model is sensitive to quantization, resulting in a significant drop in accuracy. On the other hand, forward-forward models offer the benefit of independently training each layer, allowing for more flexibility in optimizing the training process. Hebbian models were not found to be competitive, displaying performance below the required threshold. Smaller models for MCU and FPGA would likely perform even worse.

X-ray imaging systems play an important role in the diagnostic process of various medical conditions. Generating an accurate artificial X-ray image has multiple advantages. It allows for flexible configurations during generation. The resulting images can reduce testing time and cost, help the training of surgeons, and increase the amount of data for artificial intelligence model training. The generation of an X-ray image involves the simulation of a raytracing algorithm through a data model. In this research, a naive approach to this problem is examined. It was found that this approach can be improved by implementing model parallelization, data caching, and data compression. The resulting algorithm is simulated and validated in a software environment. This is then implemented for both an Ultrascale+ and a Versal FPGA. The results show that the algorithm can achieve real-time X-ray image generation, matching the performance of currently used detectors, provided that the required memory performance is achieved.

Direction of Arrival (DoA) estimation is an important topic in radar application and has significant importance for advanced driver assistant systems in the automotive industry. While there is an increasing need for higher resolution and increased target detection and DoA estimation performance, such improvements often require increased hardware cost, complexity and size. With the increase in computational power of modern systems, Compressive Sensing methods have become more attractive as alternative methods for DoA estimation to the established ones, which often rely on uniform linear arrays (ULA) and the acquisition of multiple snapshots to provide good performance. Compressive Sensing methods have been shown to fit very well into the DoA estimation framework and have the ability to use far fewer snapshots, provide super resolution capabilities and by nature, utilise sparse spatial sampling, i.e. sparse antenna arrays. The latter point is the key incentive of this thesis.

Specifically, Bayesian Compressive Sensing (BCS) which in addition to point estimates also provides measures of uncertainty is used in this thesis, to generate and use sparse
linear array structures for DoA estimation. In particular, the entropy of the recovered coefficient vector is reduced in each step. Two array generation algorithms are proposed
building on the same concept to generate sensor arrays for the consideration of a uniformly spaced, linear grid of possible sensor locations and for a Multiple In Multiple Out
(MIMO) array setup. Utilising sparse arrays with BCS has the potential to reduce the hardware complexity of the circuit board, reduce energy consumption and heat generation, as well as ultimately saving costs in production and operation.

The proposed array generation algorithms are first tested and assessed with simulated Frequency-Modulated Continuous-Waveform (FMCW) radar data, where it is shown that
the generated algorithms achieve good estimation and detection performance with a heavily reduced number of sensors compared to their fully filled template arrays. Moreover, they are shown to outperform randomly generated arrays in most cases that have been studied. To add practical insight, the generated antenna arrays are tested with measured data that has been captured in two measurement campaigns with a Texas Instruments Cascade Evaluation board, featuring an 86 element virtual ULA, which has been used as the grid of possible sensor positions for the array generations. The simulated results are affirmed by the measured data, although more sensors tend to be required depending on the clutter present in the scene. It is shown, that BCS can work very well with the proposed, heavily sparse arrays tested on both simulated and measured data, which translates directly to a possible reduction in required hardware antennas. Although in this thesis the possible sensor positions have been confined to a grid of positions spaced by half the wavelength, it is easily possible to extend the procedure to a more finely divided search space.

The primary goal of this study is to infer a time-domain (TD) counterpart of the frequency domain leaky wave. To reach this goal, a number of challenging problems will be tackled: (1) Develop adequate computational tools for performing detailed TD electromagnetic (EM) analyses via CST Studio simulations, the post-processing of the results included. (2) The TD study of the radiation from a gap-fed long slot at an air-dielectric interface. (3) The TD study of the radiation from a gap-fed long slot located just below an air-dielectric interface. The results of these explorations represent a first step towards establishing a causal counterpart of the leaky-wave radiation.

Mimetic formulations, also known as structure-preserving methods, are numerical schemes that preserve fundamental properties of the continuous differential operators at a discrete level. Additionally, they are well-known for satisfying constraints such as conservation of mass or momentum.

In the present work, a Mimetic Spectral Element Method based on quadrilaterals is explored. As an introduction, the framework is first implemented and tested on the classical Poisson equation, the Hartmann Flow system and several eigenvalue problems for the Laplacian operator. Solutions are attained by direct/mixed formulations and the extension to multi-element approaches is dealt with using either gathering or connectivity matrices. Different boundary conditions and various geometries are utilized.

Afterwards, the Maxwell Eigenvalue problem for the electric field E with general material properties is tackled in an attempt to generate spurious-free solutions by incorporating the condition ∇·D = 0 into the discrete system. The formulation is further scrutinized on geometries with Betti number b₁ > 0 as to verify if the proposed scheme captures the physical zero eigenvalues.

In the end, a mixed formulation for the eigenproblem is proposed in which the curl-curl operator is separated. The approximation of the electrostatic field energy is then computed with this formulation and compared to the solution obtained with a direct method allowing to create an upper and a lower bound for this variable.

Lens array architectures are being developed for high frequency applications such as phased arrays for 5G point-to-point communications, Fly’s Eye lens arrays for wideband wireless communications and scanning phased arrays at sub-millimeter wavelengths. Each of these applications employ different feed configurations, lens geometries and require specific feed isolation levels. In these architectures, the mutual coupling between feeds is an important design criteria. This project proposes a theoretical model based on Geometrical Optics and Physical Optics to represent the multiple reflections in a dielectric lens antenna and its impact on the mutual coupling between lens feeders. This model is then employed to study the mutual coupling in architectures with different feed models, edge illumination levels, feed positions, lens diameters and dielectric materials.

The increase in quantity and electrification of energy in residential areas pushes the existing distribution network to its limits. If left unchanged, the network will suffer from voltage violations caused by the increase of electric vehicle (EV) charging, energy generation by Photovoltaics (PV) panels and the phasing out of natural gas. Typically, cables are reinforced to cope with increased demand, but upgrading large parts of the Dutch electricity network is expensive. By not expanding the network, the remaining capacity of the network needs to be optimally used. This situation encourages the switch to an active distribution network, which opens up a new possibility for voltage management and optimal use of the distribution network.
One of these prospects is the multi-port converter. This converter combines the usually separate converters of EV, PV and battery energy storage system (BESS) to a single entity. Combined with an optimal charging algorithm, the energy management system (EMS) of the multi-port converter decides when to (dis)charge the BESS and EV, using a moving horizon window. By controlling the previously nondispatchable loads, flexibility is generated on the demand side. Using a network of multi-port converters on its own is insufficient to cope with voltage problems in the distribution network due to the overlap of (dis)charging schedules and lack of communication.
A Voltage management system (VMS) is used to coordinate the EMS of all the multi-port converters in the network to stay within the voltage range. This controller aims to prevent and correct voltage problems by varying the input or output of the multi-port converter. The prevention is done by actively reserving energy and changing prices to encourage changes in the (dis)charge schedule of the multi-port converter. The correction of the voltage profile is achieved by active power (Volt-Watt), reactive power (Volt-VAR)and apparent power (Volt-VA) control. 
Due to the stochastic nature of solar energy and load demand, the actual battery capacity and voltage profile may deviate. To combat the change in battery capacity, a reference tracker is installed. For the change in voltage, a real-time controller is implemented, used for rapid response. These controllers are placed inside the EMS under the optimal charging algorithm. 
The main focus of this thesis is to design a control structure, which can be gradually deployed, to prevent and correct voltage problems in a low voltage feeder in various time scales. By using multiple intertwined controllers, the proposed hierarchical control structure achieves fast operation, prevention and correction of the voltage profile over an extended period.
The simulations have been validated on the IEEE European low voltage test feeder. Results demonstrate that a limited number of multi-port converters is needed to keep the voltage profile within limits. The various correction methods have been tested for effectiveness to correct voltage deviations. Both active (Volt-Watt) and apparent (Volt-VA) power show capabilities of voltage control. Furthermore, the real-time controller is capable of removing individual load usage from the network and prevent voltage violations caused by uncontrolled PV and EV.

The interest for phased array antennas has increased over the past years along with the need for wide bandwidth arrays. Connected arrays are becoming more prevalent as a well suited solution to achieve the desired bandwidth. In order to increase the bandwidth and the scan angle of this type of array, horizontal artificial dielectric supersaturates have been used with good results. However, those superstrates come with an increase in cross polarization. The cross polarization of connected arrays with isotropic dielectric substrates and superstrates is thoroughly investigated in this thesis, along with an analysis on the cross polarization of the connected array with artificial dielectrics. In order to reduce the cross polarization as a result of that superstrate, a novel artificial dielectric is introduced which implements vertical vias to form a wire medium. For this medium a method of moments analysis is performed and a closed form expression is found. Additionally, the wire medium is implemented in the horizontal artificial dielectric to form the top-hat loaded wire medium as a solution to the restricted height between layers of the artificial dielectric slab. The wire medium increases only the z-component, since the wire is oriented vertically, of the effective permittivity without impacting the other effective constitutive parameters. The increase of the refraction index for transverse magnetic wave reduces the cross polarization level and aids in the matching for E-plane scanning, while maintaining a decay over the scan angle. However, a trade off between the reduction in cross polarization and the E-plane scanning is observed. This trade off is a result of the large z-component of the effective permittivity which promotes the propagation of the TM0 surface wave mode. The effectiveness of this novel structure is shown by considered a satellite communication application for simultaneous operation in the Ku- and Ka-transmit bands with a scan angle up to 60 degrees . A design for this application is performed with good matching, -10dB for broadside and -6dB for scanning in both E- and H-plane, and with a low cross polarization level of < −10dB. An evolved design with the manufacturing necessities, such as bonding layers and support layers, is obtained which exhibits a cross polarization level < −15dB with the use of the introduced wire medium, however, the layer height restriction set by the manufacturer and the feeding structure still provide a challenge at the frequency bands which are considered.

In this thesis, a closer look will be taken into the design of a propulsion system of an eVTOL (electric Vertical Take-off and Landing aircraft). The primary goal of the eVTOL is to maximize the payload for a given maximum take-off weight and the eVTOL has to go into service in 2030. To make an as light weight as possible propulsion system, an investigation will be done for the propeller length and the number of rotors, which is present on a given eVTOL design. In addition to this, different drive choices will be discussed and an optimal drive choice will be made. For this drive, a closer look will be taken in efficiency optimisation. From these parameters a design example will be given using the current state-of-the-art drive technology. In addition to this possible future drive, developments will be taken into account. 

In order to ensure safety and prevent collisions on road, automotive radars must be fault proof and have to be tested on reliability and performance, which requires proper diagnostic of well-functioning of the radar so that the car may participate to traffic. One way of the diagnostic of well-functioning of the automotive radar is by means of calibration in service stations. In contrast to offline calibration in service stations, one another way of testing the automotive radar on well-functioning would be by means of monitoring the state, or with other words the healthiness, of the radar in real-time. One possible solution to monitor the radar state is to use a massive set of calibration targets in road infrastructure and thereby the problem lies in optimally estimating the state based on the RCS information provided by the radar. As a result, in the first part of this thesis, based on the defined target selection criteria, selection of the most appropriate calibration target among possible candidates is discussed. Using massive set of targets is required to overcome the uncertainty in production and installation accuracy in one-target measurements. However, this method brings randomness which is caused by two error sources being the non-ideal shapes of the calibration targets originating from mass production errors and orientation errors either from installation or maintenance errors. Second part of the thesis investigates how this randomness affects the self-diagnostics performance of automotive radar. Thereby, the model of target orientation and RCS loss due to orientation errors, and, the model of target RCS and its RCS loss due to mass production errors are developed. For both error sources, the statistical characteristics of the loss factors are determined by means of corresponding probability distribution functions which are derived analytically. In case of orientation errors, analytical results are validated by Monte-Carlo simulations as well as Kullback-Leibler Divergence. In case of mass production errors, analytical results are validated by Monte-Carlo simulations only. Together with the results obtained in the first part, results of the statistical characteristics of the loss factor due to non-orthogonality help to find a balance between the size, quality and number of targets to be deployed in a certain range in a given road configuration. To finalize the project, the measurement model is determined according to which the approach for self-diagnostics is developed under certain assumptions and considering a set of measurements of the same realization of a single target only. By the developed approaches, the relation of statistical parameters on self-diagnostics performance is determined by means of three different estimation methods and the results are validated by Monte-Carlo simulations.

For radar imaging, resolution and dynamic range are two critical metrics to measure image quality. MIMO array exploits the spatial diversity of transmitters and receivers to achieve high cross-range resolution with fewer antennas than array-based imaging.
Generally, to make full use of the sparsity, at least the transmit array or the receive array does not satisfy the Nyquist sampling criterion. Therefore, the grating lobes may occur, which would produce ghost targets and submerge the weak targets. For far-field imaging, the grating lobes can be eliminated with various array design methods. In contrast, it cannot be completely eliminated for near-field imaging applications. Besides, the high sidelobes would also mask the weak targets and reduce the image's dynamic range. The traditional weighting method can effectively suppress the sidelobes, but the mainlobe resolution would be reduced.
In this thesis, we propose a method based on the spatially variant apodization method to enhance the image quality of near-field 1-D and 2-D MIMO array imaging. According to the generalized matched filtering imaging method, we can individually analyze the wavenumber spectrum in the cross-range and range directions of the transmit array and the receive array. Then the method can be easily implemented in the space domain to suppress the sidelobes and grating lobes without sacrificing the cross-range and range resolution. Moreover, three acceleration approaches are proposed to reduce the computation burden.
Both numerical simulation and experimental validation indicate that the proposed method can effectively suppress the sidelobes and grating lobes without spreading the mainlobe. Moreover, the acceleration methods can suppress the sidelobe and grating lobe level meanwhile decrease the processing time remarkably under some conditions.

In this paper, a method is proposed to automatically correct misalignment of orthophotos in time-series caused by an inaccurate geotransform. The proposed method relies on common literature concepts such as keypoint identification, keypoint matching, and model fitting using random sample consensus (RANSAC). Traditional keypoint identification methods such as the scale invariant feature transform (SIFT) are not suited for this problem as no real scale- or rotation-invariance is required, instead, time-invariance is required. To achieve this, crops are suggested as keypoints, and two different keypoint descriptors are put forth. The first descriptor is based on the shape and size of the crops, while the alternative descriptor is based on the planting pattern of crops. The method, and both descriptors, generate promising results for certain scenario's. However, in later growth stages performance drops significantly as the identification of crops -required beforehand- becomes troublesome due to dense growth.

The objective of the bachelor graduation project detailed in this thesis is to develop a wireless sensor patch that monitors the health of newborns. The sensor patch replaces the monitoring tasks of nurses during maternity care. The patch measures temperature, heart rate, oxygen saturation and bilirubin. Many babies are born with high bilirubin, causing a condition called newborn jaundice. All measurements will be done once every two hours and the data will be sent to the hospital.
This thesis focuses on everything related to energy and power within the wireless sensor patch. Energy storage, power and energy management and visualising data with the aid of a MATLAB model are topics that will be designed and implemented in this thesis.

The problem of quantifying the growth dynamics of plants from time-lapse images is important for determining variety-specific characteristics and subsequent breeding. The goal of this paper is to address this problem and reconstruct three-dimensional images such that growth becomes visible. To this end, data of growing plants were collected in the form of multi-angle time-lapse images. In each image separately, the plants' outline was detected with a GVF-Snakes algorithm. The success of this algorithm is ensured by pre-processing steps and placing the initial contour based on color specifics of the image.
The two-dimensional contours are then combined to reconstruct a three-dimensional contour in the camera reference frame. To obtain this reconstruction, points along the contour are matched using a RANSAC-based algorithm to estimate the essential matrix. The three-dimensional reconstructions show information about size, height, leaf structure, and growth of the plant, but the stability of the method has to be improved. Finally, an attempt is made to combine the GVF-Snakes with the point-matching algorithm, however, the convexity of the coplanarity constraint should be explored further.

For a proper three-dimensional reconstruction of histology serial sections, adjustment of the slices is necessary for combining serial sections. Deformations occur due to the sectioning and acquisition pro-cess of the microscopic analysis of histology. By reconstructing the deformations with a transformation of the sections, a mathematical correction on the images can be applied. By using image registration, a transformation function is searched for to minimize differences between the histology slices. Due to the non-linearity of the distortions, prior knowledge is required in order to have a solvable problem. Additional information in the form of elasticity regularisation is considered. Implementing the elastic regularisation with a finite element method, provides a continuous transformation function With the continuous function, in a natural way, alignment can be monitored for folding transformations. In this work, the (bi-)linear and (bi-)quadratic elements for the finite element method are implemented and compared with the finite difference method. It is observed that for the different kinds of elements, the (bi-)linear elements yield best results with the validity of the transformation. Moreover, the computa- tional costs for the bi-linear elements are the cheapest. Compared with the finite difference method, the differences in accuracy are not noteworthy but the computational time of the finite element method is longer. Furthermore, to steer the matching in an accurate direction, improvements are proposed by applying local stiffness of the elements or adding soft constraints on the volume of the elements. This results in significant improvements in the transformation. For these two approaches it is observed that local stiffness is more restrictive than volume-preserving. Solving the optimization problem, a Gauss-Newton method to search for descent directions is applied. A matrix-based and a matrix-free approach of elasticity regularisation is considered in the linear system of finding a descent direction. While the matrix-free approach decreases the memory usage, the computational costs are significantly increased.

A major cause for voiding dysfunction is the inability to relax the urethral sphincter. KiloHertz Frequency Alternating Current (KHFAC) stimulation can block signals that are travelling through the body; applying this type of stimulation at the pudendal nerve could inhibit the pulses that lead to contraction of the urethral sphincter, and restore voiding ability. In order to design a successful KHFAC block therapy for the pudendal nerve, it is necessary to understand what impact different stimulation parameters have on efficacy, safety and power-efficiency. This thesis will therefore test earlier researched KHFAC stimulation parameters against a new quality measure, study the impact of new waveform alterations, and study how bipolar electrode design can improve KHFAC therapy. By utilizing the theory behind the mechanism of the KHFAC nerve block, a new block-determination model was developed that is over thirty times faster than the classic model. The McIntyre-Richardson-Grill model was chosen as the implementation of the axon model, and the bipolar electrode was modelled as an electric dipole. The simulation experiments revealed that the charge per phase of the KHFAC signal at block threshold could be reduced, without increasing the amplitude of the signal, by introducing interphase delays to the waveforms and by creating asymmetric charge-balanced waveforms. Triangular waveforms were shown to also require less charge per phase than a regular square wave to block, albeit with a higher amplitude. A correctly aligned bipolar electrode set-up with an interpolar distance that was about the same as the electrode-to-axon distance was shown to result in reduced block thresholds. Overall, this thesis has shown how stimulation parameters can be chosen to develop an effective KHFAC block therapy for the pudendal nerve.

Image registration is a vital tool in medical image analysis with a large number of applications assisting the medical experts. Currently, conventional image registration approach with predefined dissimilarity metric and iterative optimization, is widely used. In this thesis, we proposed a method to solve medical image registration problem using a deep learning approach. A convolutional neural network architecture is proposed, named as "RegNet", applied on monomodality image registration problem. The proposed RegNet does not require any dissimilarity metric and is capable of directly estimating the spatial relationship between two images. The training is based on the pseudo-real-world displacement vector field, created by the conventional registration tool and artificial deformation simulation, resembling a deformation similar with the real-world deformation case. Multi-stage framework is also implemented to increase the capture range of RegNet. This thesis evaluates the performance of "RegNet" for an intrasubject magnetic resonance brain images registration problem.