Autonomous buses offer benefits such as fewer traffic collisions and lower emissions, but public acceptance is limited by perceived safety: the subjective feeling of being unsafe. This project explores how interior design can support perceived safety in fully autonomous buses.

Research identified three core problems. A user survey (n=98) showed that passengers feel significantly less safe without a driver (safety score drops from 4.2 to 2.8 out of 5), mainly due to "no contact", "uncertainty", and "no control". Expert interviews with HTM, EBS and RET confirmed that passenger behaviour is the biggest concern in practice. Literature showed that predictability reduces stress and that visible staff presence is the most effective safety measure, but both disappear without a driver. Through this research three core problems were identified: support (passengers need to know they can get help), social safety (passengers need to feel safe around others), and uncertainty (passengers need to understand what the bus will do).

The design framework translated these insights into four themes: Overview & Comprehensibility, Control & Calm, Social Safety, and Accessibility & Inclusivity. Through ideation, six concepts were developed and evaluated against sixteen requirements and fourteen wishes. Two concepts were combined into the Heartbeat Bus, built around two core elements. The help ring provides a central point for route information and direct contact with a help centre. The breathing ceiling makes bus behaviour visible through gentle light signals. Supporting elements include large windows, mirrors, camera screens, emergency buttons, silent alarms, QR codes, and door lights.

Validation showed that 71% of respondents rated the redesigned bus as better than a normal autonomous bus, and 83% of final test participants would feel comfortable using it without a driver. A feasibility interview with BYD confirmed technical feasibility within 15 years. The concept meets all requirements and scores well on all wishes.

This project demonstrates that interior design can significantly address the psychological and social gaps left by the removal of the driver, uncertainty of the bus and behaviour of other passengers, offering a coherent, integrated solution that makes passengers feel safe, informed, and in control.
... Read more

Accurate state of charge (SOC) estimation is essential for battery management systems (BMSs) in heavy-duty vehicles (HDVs), where batteries are exposed to high voltage and current stress, heavy loads, and dynamic operating conditions. In the context of company-oriented BMS development, the SOC estimator should be accurate, suitable for real-time embedded implementation, scalable to large battery systems, and transparent enough for engineering validation. Lithium iron phosphate (LFP) batteries are attractive in HDV applications because of their safety, relatively low cost, and long cycle life, but the long plateau region in the voltage--SOC curve limits the reliability of voltage-based SOC correction and calibration. For this reason, Coulomb Counting (CC) is selected as the basic SOC propagation framework in this thesis. CC is practical and industrially relevant because it directly uses measured current and has low computational burden, but its sensitivity to initial SOC, current measurement error, and battery capacity mismatch creates challenges for reliable BMS implementation.

The thesis develops a Simulink-based benchmark framework to evaluate selected CC-based SOC estimation and correction strategies for LFP batteries under heavy-duty vehicle operating conditions. A conventional CC estimator, two literature-based correction methods, an improved bias-compensation strategy, and a bias-capacity-aware correction method developed in this thesis are implemented and compared under controlled error cases. The results show that no single correction mechanism is optimal under all conditions: initial SOC correction depends on voltage observability, cycle-level capacity correction is more effective under capacity mismatch, and current-bias compensation is necessary when current sensor drift causes accumulated SOC error. These findings demonstrate that reliable CC enhancement for LFP-based heavy-duty battery systems requires not a stronger correction mechanism, but one matched to the dominant error source and constrained by the reliability of the available correction signals. The benchmark provides an early-stage evaluation platform for future company BMS development, supporting the selection of correction strategies before experimental validation and embedded implementation. ... Read more

This research investigated how climate change may affect characteristic values for precipitation, snow, and wind loads on structures in the Netherlands and assessed the implications for structural design. Historical observations from KNMI weather stations were analysed and compared with the current characteristic values prescribed in the Dutch National Annexes. Future changes were assessed using the KNMI'23 Climate Scenarios.

For precipitation, recent Dutch depth-duration-frequency studies indicate that the current 50-year return level for 5-minute precipitation events is already approximately 20% higher than the value prescribed in the Dutch National Annex. The climate projections indicate a further increase in extreme precipitation, resulting in an estimated factor of change of approximately 1.06 for a 50-year return level under 1.1°C global warming relative to the 1991-2020 reference period. Combined, these findings imply that emergency drainage widths may need to increase by approximately 27% compared with current practice.

Historical snow depth observations showed decreasing trends in annual maximum snow depth, while estimated 50-year return levels were generally lower than the current characteristic ground snow load. Although quantitative snow projections are unavailable, multiple climate indicators consistently suggest that conditions favourable for snowfall and persistent snow cover will become less frequent under future climate change.

For wind, estimated 50-year return levels of the 10-minute mean wind velocity were generally comparable to or lower than the values prescribed in the Dutch National Annex. Furthermore, the KNMI'23 Climate Scenarios project only minor changes in extreme wind velocities relative to the associated model uncertainty, indicating no clear need to revise the current characteristic wind loads. ... Read more

Time-lapse (4D) seismic is a critical technique for reliable monitoring of geological CO₂ storage. Its effectiveness depends on converting time-lapse seismic observations into quantitative estimates of CO2 saturation. We propose a neural-network-based framework trained with a physics-guided objective that embeds seismic and rock-physics relationships. The network adopts a U-Net-style encoder–decoder architecture with residual convolutional blocks and a Transformer bottleneck. The network predicts the non-wetting-phase saturation field from observed angle-dependent seismic reflection data. The utilized loss function compares physics-based predicted angle-dependent 4D seismic images with observed ones and also includes smoothness priors to enforce lateral continuity while suppressing spatial blurring in the estimated saturation fields. We evaluate the proposed framework on synthetically ‘observed’ time-lapse multi-angle seismic images generated from reference saturation fields produced by an 800-year flow simulation. The DNN is trained in a self-supervised manner on the first 720 years, with the remaining 80 years held out for further evaluation. The framework reproduces the large-scale evolution of two migrating plumes over the full 800-year sequence, consistently recovering plume locations, migration paths, and extents. These results indicate the framework is practical for monitoring tasks emphasizing plume localization and extent, enabling fast, stable inference of saturation changes from time-lapse multi-angle images. ... Read more

Induced seismicity related to gas production remains an important scientific and societal issue in the Netherlands, especially because earthquakes may continue after production has stopped. This post-production activity, known as trailing seismicity, creates uncertainty in seismic hazard assessment, since the largest earthquake does not necessarily occur during the production phase. This thesis investigates whether the largest observed trailing earthquakes in Dutch gas fields, with specific focus on the 2025 Zeerijp earthquake in Groningen, can be justified using an adapted framework of Ryan Schultz combined with Gutenberg-Richter magnitude-frequency statistics.
Earthquake data from the Koninklijk Nederlands Meteorologisch Instituut (KNMI) catalogue were divided into production and post-production events for selected Dutch gas fields. The main parameters used in the analysis were the magnitude of completeness, the number of production and trailing earthquakes, the production-seismicity ratio, the Gutenberg-Richter b-value, and the maximum magnitude parameter. Groningen was used as a control case because of its larger earthquake catalogue, while Annerveen, Eleveld and Roswinkel were analysed as smaller gas field examples. A Monte Carlo approach and logic-tree structure were used to include uncertainty in the input parameters and to estimate distributions of the largest expected trailing magnitude.
The results show that the observed Zeerijp earthquake with magnitude 3.4 falls within the predicted Groningen magnitude distribution. For the smaller gas fields, their observed largest trailing magnitudes are justified as well; however, uncertainty increases. Overall, the adapted framework can justify the largest observed trailing events within the selected Dutch gas fields, but the reliability of the prediction depends on the assumed b-value, completeness magnitude, seismicity ratio and maximum-magnitude constraint. Therefore, the method is useful as a probabilistic tool but should be interpreted together with field-specific geological and catalogue uncertainties. ... Read more