Automotive radar systems increasingly operate in an occupied frequency spectrum, raising concerns about mutual interference among radar sensors. Moving from the widely used 76–81 GHz frequency range towards 140 GHz provides a possible way to address this increasing interference. Operation at this frequency, however, introduces higher propagation losses and requires higher antenna gain to accommodate medium-to-long detection ranges. The resulting directional antenna responses require beam scanning to cover the required field of view, creating challenges for angular estimation under the timing constraints of automotive radar.

This thesis investigates single-snapshot direction-of-arrival estimation for beam-scanning radar systems. A simulation framework is developed to model the considered radar architecture and evaluate different estimation approaches under representative operating conditions. The methods are assessed through Monte Carlo simulations considering estimation performance, detection performance, target separability and convergence behaviour across different signal-to-noise ratios and target configurations.

The results demonstrate that the available beamspace information can support accurate single-snapshot direction-of-arrival estimation and separation of closely spaced targets. The investigated methods provide different trade-offs between estimation performance, angular resolution and computational complexity. The developed framework is not restricted to a particular antenna design and can support the investigation of different beam configurations and beam-scanning applications. ... Read more

Weather-radar networks must allocate limited sensing time between repeated observations of evolving storms and background surveillance. Fixed scanning strategies provide systematic coverage, but cannot adapt the use of radar resources when several storm observations compete for service. This thesis investigates adaptive task scheduling for a multi-radar weather sensing network through two proposed approaches: Radar-Level Task Selection (RLTS) and Load-Aware Task Scheduling (LATS).

RLTS combines Time-Balance revisit urgency with dynamic storm priority, task feasibility, and radar-dependent information to select the next observation. LATS extends this approach with explicit storm-to-radar assignment based on scan time, traverse time, urgency, and radar workload. The methods are evaluated against cyclic storm-by-storm tracking and continuous 360◦ scanning using 50 Monte Carlo trials of 20 min across five storm scenarios, simulated in a developed framework.

The results show that adaptive scheduling becomes increasingly beneficial as radar resources become constrained. LATS achieves the shortest median revisit time in four of the five scenarios, while the cyclic baseline remains strongest under low load. Under high load, LATS reduces the median revisit time by 12.9% relative to RLTS while increasing mean surveillance coverage from 3.79% to 13.13%. Tracking accuracy remains broadly comparable between the two adaptive schedulers. The results demonstrate that adaptive task selection improves storm servicing when observations compete for radar time, while load-aware assignment provides an additional benefit by distributing the workload more effectively across the radar network and preserving greater surveillance capacity ... Read more

Raw radar perception is emerging as a promising approach for all-weather autonomous driving. While current deep learning approaches have made significant efforts using established vision-based architectures, they often rely on localized spatial convolutions that do not fully exploit the underlying wave-based physics of radar signals, such as phase coherence and global frequency structures. To harness this potential, we propose WaveRadNet, an architecture that introduces two physics-aligned novelties: (1) a complex-valued MIMO pre-encoder to preserve native phase information, and (2) a hierarchical Fourier-domain encoder based on Global Filter Network to enable un-truncated global feature mixing.

Integrating these complementary modules improves the baseline model (FFTRadNet), achieving state-of-the-art performance on the RADIal dataset with an overall 95% F1 score and 81% mIoU. On the Hard split, WaveRadNet delivers a 90% F1 score, an absolute improvement of more than 11% over the baseline. When configured with a reduced number of active antenna pairs, our sub-sampled network still outperforms the full-array baseline. Finally, WaveRadNet improves perception edge cases, particularly for long-range and off-boresight targets. ... Read more

Binaural speech enhancement in hearing aids requires accurate estimates of the acoustic transfer functions (ATFs) of the active sources and the noise covariance matrix to drive spatial filters that simultaneously suppress interference and preserve binaural cues. Existing methods estimate these quantities sequentially, introducing error propagation and a dependency on noise-only intervals that limits robustness in adverse acoustic conditions. This thesis proposes a covariance-domain extension of Simultaneous Orthogonal Matching Pursuit that jointly estimates the active head-related transfer function (HRTF) subspace, source signal energies, and noise covariance from the observed cross-power spectral density matrix within a single iterative procedure. Controlled simulation experiments against a GEVD-based baseline demonstrate consistently lower HRTF estimation error across a wide range of SNR levels, time block sizes, and source azimuth configurations, with particular advantage at low SNR and short block lengths. When used to drive a joint binaural LCMV beamformer, the framework yields near-zero and stable target interaural time difference errors across the full SNR range, confirming that the joint estimation strategy produces internally consistent beamformer inputs that support robust binaural cue preservation. ... Read more

WiFi sensing enables non-intrusive, device-free monitoring of human activities by analyzing Channel State Information (CSI) extracted from commodity WiFi signals. While most research has studied Human Activity Recognition on pre-segmented clips, the harder problem of temporal activity segmentation — partitioning a continuous CSI stream into labeled activity intervals — has received less attention, and progress is limited by the absence of high-quality datasets and standardized evaluation infrastructure.
This thesis addresses that gap through three interconnected contributions. First, we introduce WiPos, a multimodal dataset in which a subject performs activities at freely varying positions, annotated with millisecond-scale precision using motion capture. Second, we present Breaking-CSI, a unified benchmarking framework that enables fair, reproducible comparison of segmentation methods across multiple datasets. Third, we propose DopplerTAS, a temporal activity segmentation model that operates on Doppler features derived from the time-differential CSI phase rather than raw amplitude, making predictions largely position-invariant.
Experiments using Breaking-CSI to evaluate representative baselines from the literature show that all of them suffer a consistent accuracy drop on WiPos compared to their native datasets, confirming that positional variation is the dominant challenge. DopplerTAS achieves 96.7% frame accuracy and 90.4% mIoU on WiPos, improving over 30 percentage points on both metrics.
Together, these contributions provide the dataset quality, evaluation thoroughness, and modeling approach needed to advance WiFi-based temporal activity segmentation from isolated recognition experiments toward continuous, position-robust sensing.
... Read more

This thesis focuses on the pipelaying tensioners of a pipelay vessel. Pipelaying tensioners maintain tension on the pipeline as it descends to the seabed and thereby prevent buckling during installation. Due to their critical role in the pipelaying process, failures can have severe economic and safety consequences. As a result, these systems are maintained conservatively, which often leads to premature maintenance actions, wasted useful lifetime, and increased costs.

This thesis investigates the feasibility of developing a data-driven health indicator to support more efficient maintenance planning decision making. A data pipeline is developed, starting from raw operational sensor data stored in a database. This pipeline transforms the data into a scalar health indicator that represents the condition of the main drive system of the pipelaying tensioner. The health indicator is derived using an autoencoder neural network that is trained exclusively on data representing healthy system operation. Deviations from the learned healthy state are quantified through the reconstruction error of the autoencoder and interpreted as degradation or anomalous behavior.

To enable practical use of the health indicator, a threshold is derived using an unsupervised approach based exclusively on healthy operational data. This threshold separates healthy and anomalous system states without using failure labels, providing interpretability to the health indicator.

The approach is validated on a test dataset containing ground truth labeled failure data. The results demonstrate that the derived health indicator, in combination with the unsupervised threshold, is capable of separating healthy from anomalous system conditions. The validation yields a classification F1-score of 0.957, showing strong detection performance and demonstrating the potential of the proposed methodology to support maintenance planning decisions. ... Read more

Autonomous driving is a rapidly growing sector that is attracting increasing attention from industry and academia. The rise of deep learning techniques has made it possible for autonomous vehicles to perceive the environment around them, including detecting objects of interest around the vehicle. The technology used to detect objects is quite dependent on sensors such as cameras, radars, and LiDARs, which are commonly used in autonomous vehicles for 3D object detection. Current state-of-the-art 3D object detection methods often use a fusion of LiDAR and camera features. As such, the perception abilities of autonomous vehicles are susceptible to sensor corruptions, whether the corruptions are caused by internal sensor faults or extreme environmental conditions. A particularly hazardous corruption occurs when a LiDAR does not record points in certain regions, which can be caused, among other reasons, by dirt accumulation on the LiDAR, a wet ground, and dark-colored objects, which can absorb LiDAR beams or reflect them away from the LiDAR. Corruptions to the LiDAR can be very dangerous as they contribute significantly to the ability of multi-modal 3D object detectors, and the robustness of multi-modal 3D object detection methods against such corruptions is understudied. In this thesis, I propose Ada-UniBEV, which detects missing laser beam returns for the LiDAR based on its geometric pattern of shooting beams, and determines a severity of corruption based on the number of missing beams. Camera features corresponding to the corrupted region of the LiDAR are weighed as more important, with the exact weight being determined as a function of the severity of corruption. The weights also vary spatially around the autonomous vehicle depending on the properties of each spatial location. Ada-UniBEV improves the 3D object detection performance over state-of-the-art UniBEV in moderate to high-severity LiDAR corruption scenarios involving missing points while maintaining the same performance on clean data. ... Read more

Occupancy grid mapping represents the surrounding environment with a discretized grid, providing information about obstacles and the drivable region using sensors such as LiDAR or radar. For automotive driving applications, these maps are central to safe autonomous navigation. While both model-driven and deep learning-based approaches exist, this thesis develops a hybrid method to estimate the occupancy grid map from point cloud data. Specifically, the proposed method builds on the pattern-coupled sparse Bayesian learning (PC-SBL) algorithm, which is well suited to the block-sparse, spatially correlated structure of automotive grids. By replacing explicit parameter updates with a lightweight convolutional neural network, the spatial correlations and sparsity profile are learned directly from the data. Based on qualitative and quantitative evaluation on LiDAR point cloud data from the nuScenes dataset, we show that the proposed approach surpasses the strong PC-SBL baseline in both accuracy and runtime. Moreover, when applied without further training to LiDAR and radar point clouds from the RADIATE dataset, it marginally outperforms PC-SBL, indicating robust cross-dataset and cross-sensor generalization. ... Read more

This thesis asks whether spiking neural networks (SNNs) and neuromorphic computing constitute a promising alternative to present-day artificial neural networks (ANNs) for autonomous space missions. Focusing on a resource- and power-constrained 1U CubeSat transiting the Van Allen radiation belts, TIENOS is a toolchain that injects radiation-inspired perturbations into trained models and records the layer-specific reactions.

The framework systematically emulates dominant soft-error mechanisms by applying (i) bit-flip faults representative of single-event upsets, (ii) additive Gaussian noise as a proxy for thermal/analog variability, and (iii) dropout-style masking to approximate transient loss or zeroing of activations. Using MNIST (frame-based) and N-MNIST (event-based) benchmarks, we compare LeNet-5–style convolutional neural networks and size-matched multilayer perceptrons with their spiking counterparts to establish an ideal software training baseline. The tool produces per-layer vulnerability profiles and robustness heatmaps across a broad range of perturbation rates, quantifies activity sparsity in SNNs, and can be used to evaluate noise-aware retraining to improve robustness without any overhead, with a path towards on-chip protections such as selective redundancy (such as triple-modular redundancy for neuron parameters), ECC and scrubbing.

Results show that fragility concentrates in a limited subset of layers depending on the fault mechanism, enabling targeted hardening with modest cost. It also indicates that noise-aware retraining improves tolerance without prohibitive accuracy loss and that SNN sparsity yields favourable energy–robustness trade-offs for bursty, event-driven sensing typical of small spacecraft. In this study, the noise-aware learned weights used for inference by the two-tinyODIN setup deliver a 15% higher accuracy for up to 2% of bit-flips in the system. Hybrid ANN–SNN pipelines could further enlarge this envelope by deploying spiking computation where sparsity is highest while retaining dense processing elsewhere, acknowledging that the scalability and baseline power of current neuromorphic platforms remain practical constraints. Overall, the methodology translates environmental assumptions for a 1U CubeSat in the Van Allen belts into actionable, layer-level design rules, providing a principled basis for space-grade, energy-efficient digital SNN accelerators and an open, extensible tool to localise and mitigate radiation-induced vulnerabilities. ... Read more

Radar (Radio Detection and Ranging) sensors are cost-efficient and robust under adverse weather conditions, making them an attractive component in modern automated driving perception systems, but they provide significantly sparser information about the environment than camera or LiDAR sensors. Thus, to fully exploit radars in perception solutions, it is crucial to exploit not only local but also global contextual information of the scene. However, existing 4D radar models often struggle to fully exploit both types of information, resulting in suboptimal performance. This paper proposes DRIFT, a dual-representation model that effectively captures and fuses both local and global contexts through a dual-path architecture. The model incorporates a point path to aggregate fine-grained local features and a pillar path to encode coarse-grained global features. These two parallel paths are inter-fused via novel feature-sharing layers at multiple stages, enabling full utilization of both representations. DRIFT is evaluated on the widely used View-of-Delft (VoD) dataset and an internal dataset, demonstrating its state-of-the-art performance across multiple tasks, including object detection and free-road segmentation. Notably, DRIFT achieves a mean average precision (mAP) of 52.6% (compared to 45.4% from the CenterPoint baseline) on the VoD dataset, surpassing existing methods. ... Read more

Cardiac cine MRI is a modality used to visualize the beating heart by acquiring a sequence of images within a short acquisition window. Even if data collection spans multiple heartbeats, the goal is to capture a single cardiac cycle, that is, one full heartbeat. Achieving this often requires undersampling of k-space, which leads to artifacts that degrade image quality, such as aliasing, blurring, and ghosting. This thesis investigates how deep learning can be used to reconstruct high-quality cardiac cine MR images, which are anatomically accurate and diagnostically useful. A deep learning-based reconstruction method is developed, with particular attention given to the role of training data, network architecture, and k-space sampling strategies. The models are trained using two types of datasets, each of which is transformed to mimic cardiac cine MRI acquisition characteristics and augmented to increase the data variability. The proposed model architecture consists of initial image formation, data consistency, and convolutional denoiser blocks. The denoiser block includes variations in the use and size of temporal kernels to evaluate their impact on reconstruction performance. Experiments are conducted across 4-shot and single-beat acquisition protocols, under different sampling conditions. The proposed models achieve improved reconstruction quality over the benchmark model in most scenarios. This improvement is primarily attributed to the temporal processing of the training data and the inclusion of temporal convolutions. The findings of this work show possible paths for future research in deep learning-based cardiac cine MRI reconstruction. ... Read more

Magnetic Resonance Imaging (MRI) is a powerful tool for visualizing internal body structures and is widely used in clinical fields. However, MRI's long scanning times and high computational demands for post-processing pose challenges, especially in resource-limited environments. Recent advancements in machine learning, specifically model compression techniques, have offered solutions to accelerate MRI post-processing and make it more accessible.

This thesis systematically investigates the application of several model compression methods, such as low-rank factorization, knowledge distillation, and quantization, to enhance the efficiency of a baseline MR reconstruction neural network. By exploring multiple variations within each compression technique, this study evaluates their impact on key performance metrics such as inference speed, model size reduction, and reconstruction accuracy. Extensive tests show significant trade-offs between image fidelity and computational efficiency, providing insights into the practical feasibility of deploying compressed models in clinical workflows.

Among the techniques tested, low-rank factorization implemented via Tucker decomposition emerged as the most effective approach. This method achieved a threefold reduction in inference time while maintaining high reconstruction quality, highlighting its potential to improve MRI processing times in real-world applications significantly. ... Read more

As the Internet of Things (IoT) continues to expand across various industries, the demand for self-generating and sustainable power solutions is increasing more urgent. Piezoelectric energy harvesters offer a promising solution as they can generate electrical energy from ambient mechanical vibrations and operate independently even in locations that are difficult to reach. This thesis explores the modeling and measurements of a piezoelectric harvesting (PEH) system designed to power a low-power IoT circuit embedded in aircraft wings. Focusing on a piezoelectric cantilever beam implementation made from a PZT-5H transducer, the research evaluates three commercial models developed by 'Mide Technology': PPA-1021, PPA-2011 and PPA-4011. Every model is evaluated for their performance under mechanical vibrations in the 30-120 Hz range, characteristic for aircraft environments. The methodology includes electromechanical modeling and a structured series of laboratory experiments to validate key parameters such as resonance frequency tuning, impedance matching, power output, and behavior in noisy environments. Among the tested models, the PPA-2011 demonstrated the best overall performance achieving a peak power output of 1.61 mW at resonance of 58 Hz and was able to stably power an IoT circuit even in a noisy environment. The results validate the ability to use PEHs for the power generation of aircraft wireless sensor networks. Future work should explore long-term stability and extensive testing of real-world aircraft vibrations. ... Read more

Automotive radar is an important sensor technology for self-driving cars and Advanced Driver-Assistance Systems (ADAS). Current automotive radars lack the ability to classify and categorize objects due to their limited angular resolution. A new generation of automotive radar systems, known as automotive imaging radars, proposes to overcome this limitation by using larger apertures with more antenna elements. During the radar operation, automotive imaging radar systems face challenges in the accurate estimation of the Direction-of-Arrival (DoA) due to phase incoherency in the spatially sampled information caused by hardware imperfections, temperature variations, and aging effects. This work proposes a method combining convex optimization and alternating updates to first jointly calibrate the phase incoherencies and estimate the DoAs, and then update them iteratively. It further derives the Cramér-Rao Bound (CRB) and investigates the impact of the phase incoherency on DoA estimation using the CRB. In the proposed approach, the MUltiple SIgnal Classification (MUSIC) algorithm is applied to estimate the DoAs after each calibration. Additionally, the eigenvalue decomposition process in MUSIC is replaced by the Projection Approximation Subspace Tracking (PAST) algorithm to reduce computational complexity while maintaining the accuracy of DoA estimation. Experimental results illustrate the effectiveness of these techniques, highlighting their potential in improving next-generation automotive radar systems. ... Read more

The rapid development of Advanced Driver Assistance Systems (ADAS) necessitates enhanced performance in automotive radar systems, with Phase Modulated ContinuousWave (PMCW) radar emerging as a key technology due to its high resolution, interference resistance, and robust performance. Despite these advantages, PMCW radar faces challenges such as high computational complexity and Doppler-induced range sidelobes. This thesis addresses these challenges by proposing an adaptive block FFT correlation method to reduce computational complexity and enhance processing efficiency, ensuring reliable target detection. Additionally, we tackle Dopplerinduced range sidelobes by introducing code diversity and novel frame designs for MIMO systems, leveraging cyclic shifts and Hadamard matrices to balance sidelobe attenuation and sequence set size requirements. Through extensive analysis and simulations, the proposed methods demonstrate significant improvements in radar performance, especially in detecting weak targets behind strong reflectors. The findings contribute to developing more efficient, reliable, and scalable PMCW radar systems for advanced automotive applications. ... Read more

This thesis investigates the application of multi-agent reinforcement learning (MARL) to the optimization of radar waveforms. Radar technology is crucial in fields such as aviation, maritime navigation, and defense, but faces challenges such as interference, clutter, and the need for high resolution and accuracy. Cognitive radar, which adapts to environmental changes in real-time, offers a promising solution. This research aims to explore the potential of MARL in optimizing radar waveforms and examines whether incorporating domain knowledge can enhance performance.

The radar waveform optimization problem is framed within the Decentralized Partially Observable Markov Decision Process (Dec-POMDP) framework, defining the radar environment, agents' observations and actions, and reward functions. The study experiments with different architectures, including decentralized actors with a centralized critic. The centralized critic, having access to global state information, helps stabilize the learning process and mitigate non-stationarity and credit assignment problems. The use of GNNs as a centralized critic is proposed to leverage graph data sparsity, enhancing scalability.

The proposed models are trained and tested in a radar-tracking scenario, evaluated in terms of Pareto optimality and optimization times. The results show that both Independent Actor-Critic (IAC) and Independent Actor with Centralized Critic (IACC) models outperform traditional methods in terms of probability of detection, waveform duration, and optimization speed. The findings highlight the effectiveness of MARL approaches in optimizing radar waveforms, emphasizing the benefits of centralized critics for robustness and coordination. However, the choice of architecture significantly impacts performance, and while GNNs offer potential scalability advantages, their integration of domain knowledge did not yield significant improvements in this study. This research lays a foundation for future exploration of MARL and GNNs in radar waveform optimization. ... Read more

This thesis details the theoretical background and development process of a classification model for electroencephalogram-based (EEG) motor imagery (MI) signals, to be used in a brain-computer interface (BCI) system. This project was undertaken in order to demonstrate the possibility of distinguishing MI-EEG signals acquired using the g.tec Unicorn Hybrid Black EEG measurement cap. The classification model devised in this thesis is a hybrid deep neural network model, which combines a convolutional neural network (CNN) and long short-term memory (LSTM) recurrent neural network (RNN) in parallel, closing with a fully-connected (FC) layer. Much experimentation and research was needed to create this model, and this is extensively discussed within the thesis. The classification model produced for this thesis is one part of a complete end-to-end BCI system, which also entails a measurement and data processing procedure, and the development of a graphical user interface which provides visual feedback to the user.

On publicly available datasets, the classification model produced promising performance, achieving 55% on the BCI Competition IV-2a (4 class) dataset and 78% on the BCI IV-2a (2 class) dataset. The model has not yet been tested on self-collected data. ... Read more

Radar technology has evolved into a versatile and robust tool for critical air traffic control, meteorology, surveillance, and defence applications. In surveillance radar, the need for continuous monitoring of large areas, often cluttered by ground or sea reflections, presents significant challenges for multitarget detection. This clutter can obscure true targets, complicating detection in environments where standard radar noise assumptions fall short.

This thesis introduces a novel implementation based on the relevance vector machine (RVM) to address the complexities of multitarget detection in cluttered environments. Unlike conventional approaches that assume white Gaussian noise, the proposed method jointly estimates a clutter covariance matrix, allowing it to adapt to the estimated clutter model over subsequent iterations. Performance evaluations using simulated data in one-dimensional (range or angle) and two-dimensional (range-angle) settings demonstrate that the framework achieves accurate AR parameter estimation. Results indicate a marked improvement in reducing false and missed detections compared to the white-noise-based model. Notably, the framework performs multitarget detection without prior knowledge of target locations and the need for guard cells, underscoring its adaptability to real-world scenarios. ... Read more

In this study, we perform human identification using accumulated radar point clouds in an outdoor scene. We employ PointNet as classification network and explore the impact of adding radars' non-spatial features as input, namely doppler velocity and radar cross section (RCS). Furthermore, we encode time as an additional time identity dimension to each point within the accumulated point cloud. We examine the effects of normalizing the RCS values, canonicalizing the spatial dimensions of the point cloud, as well as normalizing the doppler velocity with respect to this canonicalization. We examine three different PointNet configurations to understand the impact of the TransformNet blocks (T-Net) within the PointNet architecture on our six-dimensional radar data input. We have created a realistic outdoor dataset for training and evaluation purposes. Our approach of using the unnormalized six-dimensional radar data on the PointNet architecture without the two T-Net blocks achieves the highest performance of 73.4 % on our test set. ... Read more

Ambiguities are an often encountered nuisance in signal processing and are the source of some of the fundamental trade-offs encountered in radar systems. The goal of this thesis is to extract unambiguous information about targets by combining a limited amount of measurements on a video integration level. A novel framework is proposed to reach this goal. At the heart of the framework lives a relevance vector machine which is extended to process the ambiguities on a video integration level and to work off-grid. The relevance vector machine is then extended to become the ambiguity aware relevance vector machine. This extension is either performed by a frequentist test or by estimating a posterior distribution. The frequentist test is used to test whether we can statistically significantly discern the returned output from ambiguities. The posterior is estimated according to Bayes’ theorem and thus allows for the incorporation of prior information. In this thesis, the framework is specifically applied to Doppler processing of a pulse-Doppler radar system. Compared to existing methods for estimating unambiguous Doppler velocity in a multi-target environment, the framework provides a general increase in performance, allows for the incorporation of prior information, and is able to give a measure of confidence in the estimates. A simulation study is set up to show the performance increase. This simulation study also highlights the utility of incorporating prior information and the quantification of uncertainty. ... Read more