This research aims at determining a suitable order release strategy for a warehouse featuring a shuttle-based storage and retrieval system with multiple goods-to-person picking workstations. These picking workstations facilitate efficient picking; however, current release strategies do not take similarity between orders into account, while this could potentially lead to better capacity utilisation. A meta-heuristic approach has been developed in literature to exploit this similarity, thereby using less stock retrievals to fulfill customer orders. However, its capabilities are limited. In this thesis, the meta-heuristic is extended to deal with multiple pick stations and a stock-multiplicity constraint. The algorithm consists of multiple steps. First, a random-start greedy algorithm generates an initial solution for a large set of orders. However, this initial solution will be of poor quality in general. As the computation cost scales exponentially with the order set size, heuristically optimising the entire order set is infeasible in practice. Secondly, the initial order sequence is split into subsets of equal length, each of which will be picked by a separate picking workstation. The picking workstations are then sequentially optimised using the meta-heuristic approach assuming an infinite stock. Afterwards, a check on the stock-multiplicity constraint is conducted, i.e. it is checked if the warehouse is capable of supplying all picking stations simultaneously. If an order line violates this constraint, the order is selected and put in a random position of the order completion sequence of that picking station. It was found that this method is capable of avoiding violations of the stock-multiplicity constraints, although it was also found that this constraint will hardly be violated in a realistic goods-to-person warehouse setting, i.e. with a large number of orders, picking stations, and order lines per order. Furthermore, our simulations show that the algorithm is capable of significantly reducing the number of stock retrievals per picking workstation.

This thesis report proposes a framework to implement Navigation, Guidance and Control (GNC) systems, that enable point-to-point autonomy for displacement vessels. A model-based control approach is chosen as the basis of the GNC systems. The resulting algorithms are implemented for verification in a 1:25 scale model of a Azimuth Stern Drive (ASD) 3111 Damen tug named "Damen Autonomous Ship", aka DASh. First, a compact maneuvering model that captures relevant dynamics of displacement vessel is formulated and identified using system identification. Secondly,the guidance system is automated such that it connects an initial state to a goal state with a collision free path that satisfies all input and differential constraints of the vessel model. To this end, the kinodynamic Rapidly-exploring Random Tree (RRT) algorithm is extended to use a maneuver automaton and optimal motion primitives in its steering function. A learned cost-to-go distance metric for the state space is formulatedto efficiently calculate distance between states, which is used to search for nearest neighbors in the kino- dynamic RRT algorithm. The performance of the planner using the learned cost-to-go distance metric is compared to a minimal curve length distance metric based on Dubins Curves and the commonly used straight-line Euclidean distance metric. It is shown that the learned cost-to-go and the minimal curve length distance metric result in paths of similar performance while the Euclidean metric performs severely worse. Lastly, the navigation and control systems are implemented on DASh. Due to disturbances present in real world environments, the paths must be tracked using feedback control. State estimation for navigation, based on position and heading measurements is performed by implementing an observer using a Extended Kalman Filter (EKF). Non-linear model predictive control (NMPC) in combination with thrust allocation is used to control the vessel during path execution. Due to real time requirements of DASh, the EKF, NMPC and the thrust allocation algorithm are directly implemented inefficient C ++, on the on-board computer of DASh. It is shown that NMPC converges to static reference positions faster and with less control inputs compared to traditional non-linear PD control. It is also shown that time-varying trajectories created by the kinodynamic RRT can be executed successfully. This shows that the identified model is suitable for use in model-based control and that the planned paths indeed satisfy the input and differential constraints of the vessel.

Max-plus-linear (MPL) systems are systems that are linear in max-plus algebra. A generalization of these systems are Max-Min-Plus-Scaling (MMPS) systems. Next to maximization and addition (plus), MMPS systems use the operations minimization and scaling. They are discrete-event (DE) systems, which means that the changing of the states is triggered by the occurrence of events and (part of) the states in the state vector represent time instances. One way to control MMPS systems is by using Model predictive control (MPC). This is a powerful on-line control strategy that uses a receding horizon. However, an efficient control procedure that works for all time-invariant DE MMPS systems had not yet been described. The goal of this master thesis is to fully design such a framework. To achieve this, the state vector is altered, such that the difference in the states that represent a time instance is included as well. Next to this, the MPC problem on an MMPS system is altered to a Mixed integer quadratic programming (MIQP) problem, in order to optimize it more efficiently. That this framework works is supported by a stability analysis. Next to that, it is tested on a simulation example of an urban railway line. Based on this example, it is shown that the procedure does indeed work. The thesis ends with several suggestions for future research.

This thesis presents modeling and control of novel tensegrity joint based on Twisted and Coiled Polymer Muscle (TCPM) that is biologically inspired by the human elbow. The tensegrity principle that is used to construct this joint has made it structurally compliance. Therefore, this joint can be categorized as a compliance actuator. The modeling is performed by constraining the joint movement to 2 Degree of Freedom (DoF). As a result, the joint kinematics and dynamics can be simplified and approximated as rigid body movement. To actuate the joint, the length of the TCPMs can be controlled individually by Joule heating. This joint can be seen as a MIMO system subject to multiple constraints. These constraints are imposed by the TCPM and used to prevent the joint from failure during actuation. Model Predictive Control (MPC) is designed to control the joint movement. This controller is suitable to handle a multivariable system subject to constraints. MPC controller based on linearized model and Hammerstein-Wiener model are designed and their performance on controlling the joint is compared. Finally, trajectory tracking simulation with various cases is provided to assess the controller performance.

Extensive advancements in the field of mobility and safety in transportation in last few decades have acted as a catalyst for the expansion of automobile industry. The research for the evolution of new technologies, vehicle safety regulations and specifications and the market requirements has given this progress a greater push. The passenger cars these days are equipped with ample collection of active safety systems such as Anti-lock Braking Systems (ABS), Electronic Stability Program (ESP), Active Front Steering (AFS) and many more. The amalgamation of these active safety systems in the vehicle chassis allows the vehicle to accomplish the desired stability. Both academia and industry understood the requirement of more research in the field of vehicle industry. This leads to the comprehensive enhancement in the field of research and development with respect to integrated chassis control (ICC). The new research delivered some remarkable advantages to this industry, such as cost reduction of the complete system, multi-objective performance growth, hardware entanglement, fault-tolerant capability and system plausibility. Despite the huge diversity in research of integrated chassis, there are certain adaptive design methods which can be used to enhance the ICC performance. This dissertation intends to provide a methodology to fully exploit the nonlinear dynamics of the vehicle concerning the forces in between tire and road surface and to discuss the integration of AFS and ESP. It establishes a novel adaptive control approach in order to attain asymptotic tracking for switched affine vehicle models with parameter uncertainties and disturbances acting on the model. This, therefore, will help to attain the desired performance. Further, comparisonwith strategies that merely exploits the linear region of the vehicle dynamics are discussed, and performance improvements of the proposed methodology are assessed. Finally, analytically the endurance of the prospective control strategy is proven and simulations are conducted to validate the theoretical analysis.

The Free Energy principle represents a Neuroscience theory that unlike any other theory can explain all behavioral aspects of adaptive agents (like humans). The issue of behavior is made computationally tractable by considering behavioral processes as various optimization problems with a single objective: minimization of Free Energy. The rationale is that to exist adaptive agents have to resist a natural tendency to disorder. This is accomplished by adopting a policy that minimizes Surprisal. To put it differently the adaptive agent has to become an expert at predicting its own sensations (eyes and proprioception) caused by an uncertain environment. The brain is therefore believed to embody a generative model. Surprisal is the measure of how good this generative model can explain the sensations experienced by adaptive agent. The mathematical equivalent of Surprisal is the negative log Likelihood. The Likelihood distribution -a typical term used in probability theory- expresses how probable the observed sensory data -sensations- is for different settings of causes. By maximization of the Likelihood Surprisal is minimized. To obtain an exact maximum Likelihood solution one however has to integrate over all possible settings of all known and unknown causes. Rather than minimization of Surprisal a minimization of a bound on Surprisal: the Free Energy is considered. The Free Energy principle formulated into a Bio-inspired Control algorithm could be the solution to the lack of adaptability and precision towards uncertain and unknown environments that the robotic community currently faces. This formulation is however not trivial. This thesis addresses the process of Learning a model of a Process according to the principles of Free Energy. It is identified the Free Energy principle to heavenly rely upon concepts from the Machine Learning methodology Expectation Maximization. In case of linearity and a Wiener process assumption on the noise affecting the Process, the Free Energy Learning algorithm is equivalent to the EM algorithm. An evaluation of the minimization of Free Energy problem with respect to the state led to a state estimation computation equivalent to the Kalman filter. An evaluation of the minimization of the Free Energy problem with respect to the model parameters led to a model parameter computation equivalent to the solution of a linear least square problem. It has been identified the EM algorithm to be quite unfamiliar in Control settings. Implementations of the Free Energy Learning algorithm = EM algorithm yielded promising results although only a naive version has been implemented. Improvements on the implementations are believed to generate a competitive algorithm for solving control problems. Next steps are consideration of the full Free Energy principle algorithm where the Wiener assumption on noise does not apply and the input is derived with a similar Free Energy minimization problem. The former could relate to the concept of incorporating a prediction horizon suggesting a intimate relationship with the Subspace Identification method. The latter is inherently different from conventional control. The full Free Energy principle algorithm does distinguishes itself from the EM algorithm.

Disturbances are an important factor in the performance of complex large scale systems. These complex large scale systems can be modeled by a broad class of hybrid dynamical systems. The current practice of controlling such processes is one way, from the scheduling level to the control level. The performance of these complex systems can be improved by allowing an exchange of information between both the scheduling and control level. Although the techniques derived in this thesis are applicable to general class of hybrid systems, this thesis focuses on a railway network. This network, or a part of it, is modeled so that information at regular time instances can be sent to a dynamic scheduler which outputs an optimal schedule back to the control level, based on the current conditions in the network. To this end, the railway system is abstracted into a Discrete-event system (DES). Essentially, the abstraction removes the continuous dynamics of a hybrid dynamical model and replaces them with discrete-events. All these discrete-events together form the railway network as a DES. Furthermore, the DES is modeled in max-plus algebra such that a Max-plus linear (MPL) is obtained. The scheduling problem is solved by a Model predictive control (MPC) scheme which is solved by a Mixed-integer linear programming (MILP) problem. For this scheme, the MPL model is converted into a Switching max-plus linear (SMPL) model which is written as a set of mixed-integer linear constraints. In contrast to the scheduler which only contains a macroscopic model of the railway network, the trajectory planner includes microscopic constraints. Therefore, to complete the scheduling, the schedule found by the scheduler is iterated between the scheduler and a trajectory planner for further improvements. A part of a railway network is studied in a case study, showing that updating the schedule results in less delay in case a delay is introduced.

In this thesis, an economic-engineering model of the ripple effect in the housing market is put forward. The ripple effect in the housing market is modeled by integrating into one model the local price dynamics of the rental market, the spatial price dynamics between local rental markets, and the valuation of real estate. Such a model is lacking in current housing-market literature. The economic-engineering model design has two main aspects. First, economic-engineering model design uses analogs between mechanical systems and the housing market, ensuring that the model contains only economically and mechanically interpretable parameters. In this way input optimization and parameter optimization have a direct implementation in the real world. Second, economic-engineering model design relies on classical-mechanical modeling techniques that make the model suitable for optimally and robustly determining governmental policy using control formalism. This thesis takes the perspective of the housing market as a heterogeneous economic space, where the dimensions are geographic proximity and economic influence. The model put forward in this thesis spatially discretizes this heterogeneous economic space into homogeneous local markets, such that local price dynamics govern the local markets and inter-local differences add spatial price dynamics. The model put forward in this thesis is designed to predict where shortages will arise due to the introduction of rent control. Such a model guides policymakers in where to build for effectively relieving shortages in the housing market. Finally, such a model informs investors about the value change to expect due to policy changes and migration trends.

Automated Case Picking (ACP) is a concept for warehouse automation made by Vanderlande Industries. A subsystem of the ACP concept is the automatic palletiser that is responsible for handling cases from a tray onto an order carrier. The research questions concern whether the automatic palletiser can be modelled as a Switching Stochastic Max-Plus Linear (SSMPL) model and what conclusions can be drawn on capacity, bottleneck and sensitivity issues. The physical components of the automatic palletiser that are handling the cases are in succession a conveyor system, a tray lift, a Tray Unloading Robot (TUR), a PickUp Table (PUT), a Palletiser Robot (PR) and a carrier lift. These physical components and several software and vision based components are depending on each other and together determine the behaviour of the system. The automatic palletiser is treated as a Discrete Event Dynamic System (DEDS) and is linear in the max-plus algebra as it contains synchronizations and no concurrency. Structural changes within the system are solved by describing it as a Switching Max-Plus Linear (SMPL) system. The system switches between modes that describe a certain system structure. The automatic palletiser is described by an SSMPL model in which stochastic parameters are used to describe the actions that take a variable amount of time. The model is written in a state-space representation which makes it suitable for implementation into MATLAB. Timing variables are based on simulation data performed by Vanderlande. Simulation of the model results in a structural difference with the reference simulation as a result of the stochastic characteristics. Cross-validation shows that the model gives the expected result when the identification and the test data are different. Analyses on activeness of the robots, cycle times and on critical cycles conclude on the TUR being the system’s bottleneck. It is recommended to research the change in accuracy due to fastening the robot. Research on Max-Plus Scaling (MPS) functions matrix multiplication makes it possible to easily describe a sequence of max-plus matrix multiplications in terms of MPS functions. This is useful in simulation applications as performed in this thesis. However, due to characteristics of the internal cycle times of the system, the automatic palletiser in the present form does not qualify for implementation of MPS functions.

Max-plus algebra is an algebra that is entirely based on the mathematical operations max(a,b) and a+b, hence the name max-plus algebra. It can be used to describe Discrete Event Systems (DES) that require scheduling, such as a printer or train network. Max-plus algebra is studied because of its interesting properties, which make some non-linear systems in conventional algebra linear in max-plus algebra. These systems are called Max-Plus Linear (MPL) systems. This exploratory study introduces Max-Plus Linear Parameter Varying (MP-LPV) systems, systems that are not entirely linear in max-plus algebra but not that non-linear either, like LPV systems in conventional algebra. An urban railway line will be taken as an example of an MP-LPV system. Urban railway lines often operate relatively freely, with a passenger-dependent variable dwell time which can be modelled in max-plus algebra as a linear variable dependency on the arrival and departure times. It will be shown that such an MP-LPV system of an urban railway line can be rewritten to a set of linear inequalities, which can be used in an optimization framework to optimize for both minimal total passenger travel time and minimal absolute operation time. Some study cases will be shown in with it can be observed that these systems compute very rapidly, which makes a possible practical implementation interesting. Finally, some algebraic analysis on the MP-LPV system of an urban railway line will be done, such as on the definition of stability. But future work is still necessary on further analysis on the general class of MP-LPV systems.

This research presents a framework for analysing the stability and control of discrete-event systems, specifically emphasising max-min-plus (MMP) and max-min-min-plus-scaling (MMPS) systems. These systems are valuable modelling tools for various applications, including production systems and urban railway traffic management, respectively. However, a critical challenge in discrete-event systems is the lack of a generalised approach to assessing the stability of time signals, particularly in the context of MMPS systems. To address this challenge, this research will use max-plus Lyapunov functions already used to study the buffer stability in discrete-event switching-max-plus-linear (SMPL) systems. This thesis provides a framework to use max-plus Lyapunov functions to determine buffer stability of MMP and MMPS systems, focusing on their time signals. The max-plus Lyapunov function uses a buffer for each pair of states. The system is considered stable if the difference converges to the buffer levels for every pair of states. Given the structure of MMP and MMPS systems, the difference between the states after one state update will often be bounded. To determine this boundedness of the buffer levels, a novel concept of "fully correlated" MMP and MMPS systems is introduced. Using the properties of fully correlated systems, an algorithm is proposed to determine the buffer levels for both MMP and MMPS systems. We also derive analytical methods using Markov properties to assess the additive eigenvalue of fully correlated time-invariant monotonic MMPS systems. Using the property of fully correlatedness, it is also derived that fully correlated time-invariant non-monotonic MMPS systems will always have a bounded buffer and growth rate and can have multiple additive eigenvalues. The findings show that fully correlated time-invariant systems consistently exhibit bounded growth rates. In addition to providing theoretical insights, this study demonstrates the practical use of max-plus Lyapunov functions as a control Lyapunov function (CLF) in model predictive control (MPC). A novel control technique is proposed to stabilise naturally unstable discrete event systems. This approach has been effectively applied to stabilise inherently unstable discrete-event max-plus-linear (MPL) and MMP systems, indicating the practical significance of the proposed framework.

Inland waterways form a natural network infrastructure with the capacity for waterborne transport of people and goods for moving freight from seaports to the hinterland. Recently, Inland Waterway Transport (IWT) has been promoted more extensively by the European Union and various governments as it plays a crucial role in reducing road congestion and CO2 emissions from transport. However, the advantages of IWT are not fully exploited due to inefficiencies in the logistics system, such as long waiting times at locks and sub-optimal navigation on waterways. Currently, no scheduling at infrastructures or routing optimisation of the overall waterway network is happening. The scheduling of vessels through a lock is usually performed on a First In First Out basis, providing an opportunity for improvement. Hence, this thesis aims to design a scheduling strategy for generating an optimal plan for sending inland vessels through a waterway network with minimal delays, yielding a significant positive impact on the modal shift towards IWT.  A promising approach to scheduling problems is by using Switching Max-Plus-Linear (SMPL) systems. SMPL systems have proven to be effective in various Discrete-Event Systems and transportation networks. Using SMPL models is convenient since non-linear scheduling problems can be described linearly using Max-Plus operators without compromising on the system dynamics. Moreover, as the SMPL systems can be transformed into Mixed-Integer-Linear-Programming (MILP) problems, it is also possible to use fast optimisers for solving the scheduling problems. This thesis will show how one can describe IWT systems, consisting of; waterways, vessels and locks, as SMPL systems. The optimal schedule for the inland vessels is determined based on multiple input parameters, including waterway network lay-out, the sailing speeds of vessels and arrival deadlines of the vessels. The scheduler will return the individual vessel routing and overall vessel order in the waterway network. This routing and order selection is defined using binary control variables, turning the IWT scheduling problem into a MILP problem, which will allow finding the solution to large scale IWT scheduling problems in a reasonable computation time. Furthermore, this thesis will show how the goal of minimising the cumulative arrival times of all vessels in a network can be achieved. This is done for different types of waterway network cases, for which the results are shown and analysed.

Legged locomotion is a discrete event system (DES) due to the ever changing contact states of each leg. As such, it requires a nonlinear modelling method to predict the trajectory of a legged robot. One such robot has been focused on throughout this thesis; the six legged "Zebro''. With its half-circular legs, the Zebro is well suited for traversing uneven terrain and even climbing up small steps, making it the robot of choice for a lunar mission in the near future. A previous attempt at modelling the trajectory of a Zebro kinematically fell short when it came to modelling a curved path, with the reasoning behind this being how the Zebro's legs can visibly be seen slipping over the ground as it walks, the effect of which is never taken into account. A combination of kinematics and dynamics was used for the model in this thesis. The reason for this is that the Zebro's legs are actuated using position controlled motors, so the information available is the leg's angle and the speed at which the leg is rotating, rather than the torque. Therefore, kinematics were used to estimate the vertical motion of the Zebro due to the rotational speed of each leg, which could then be used to estimate the normal contact force on each leg. The traction forces could then be estimated using the normal force and a slippage model which has been experimentally identified in collaboration with a different research team at the TU Delft. With these forces, and the Newton-Euler equations of motion, the Zebro's path could be predicted. Applying a slippage model to a half-circular leg, rather than a wheel, required a new method of calculating the slip ratios which was based on Pacejka's formula, but adapted to also account for the case where a leg is standing on its toe. Furthermore, a contact detection algorithm was designed to kinematically predict which leg(s) lift off of the ground during a touchdown event. This was used to kinematically model the orientation of the Zebro during a tetrapod gait and was shown to correctly predict the complicated contact transitions during said gait. Another product of the research in this thesis is a new and improved algorithm for a turning tripod gait, which achieves smoother turning than beforehand by guaranteeing a smooth contact transition between two leg groups. That being said, it cannot yet be applied to the tetrapod gait, so the current gait scheduling algorithm is still required for a turning tetrapod gait. The results of the simulations showed the Zebro walking as expected, both in a straight line and when turning, but the simulations could not be validated quantitatively due to current events regarding COVID-19. For a qualitative review of the model, photographs were taken of the Zebro during walking gaits to compare to the simulations and showed that, while straightforward walking was predicted well, the turning circle of the model was significantly sharper than in reality. The reason for this is most likely a problem in the calculation of the slip ratios, since they were consistently unrealistically low, therefore requiring further research in future.

Legged locomotion is an example of a discrete event system (DES) with only synchronization and no concurrency. This particular kind of DES can be framed in max-plus algebra, which consists of maximization and addition as its basic operations. Utilizing the max-plus operations between matrices, max-plus linear (MPL) system is constructed. Switching max-plus linear (SMPL) system is a class of MPL system where the element of the system matrices can be changed by a switching function. In modeling legged locomotion as an SMPL system, the touchdown and lift-off time instant of every leg is set as the states. The SMPL system is then used to schedule when each leg should touchdown and lift-off the ground. The swing period of a leg starts with lift-off and ends with touchdown. The stance period of a leg starts with touchdown and ends with lift-off. The SMPL system then incorporates the gait and the desired swing and stance period, to schedule the touchdown and lift-off time instant of every leg. The legged robot which is considered in is called Zebro. Zebro is a robot with six half-circular legs, developed in TU Delft. In this report, the trajectory-following SMPL system for Zebro is proposed. With the trajectory-following SMPL system, a Zebro is able to move forward in all directions by means of the legs scheduling. To turn, Zebro needs to have different linear velocity on the left and right side. While the Zebro is turning with the trajectory-following SMPL system, the stance period of the inner legs are longer by tp seconds compared to the outer legs. On the other hand, the swing period inner legs are shorter by tp seconds compared the outer legs. The motion of the Zebro is predicted by the switching kinematic algorithm in this report. Based on the relation between coordinate frames, first Jacobian matrices of the legs are formed. Each jacobian matrix describes the relationship between the motion of a leg and the motion of Zebro. The jacobian matrices are combined in the sensed forward solution, to predict the motion of Zebro if there are several legs on stance period. The switching kinematic algorithm will detect which legs are on the stance period, and the change of angle of the respective legs. The switching kinematic algorithm then forms a suitable sensed forward solution. The results of the simulation show that the Zebro can turn by the trajectory-following SMPL system. In general, if the tp is increased, the turning radius of Zebro is smaller. However, the Zebro does not always turn to the intended direction. To predict the trajectory of Zebro, a switching kinematic algorithm can be used. The switching kinematic algorithm can predict the trajectory of Zebro if a certain lower boundary of tp is exceeded. The Zebro can also turn to the intended direction if that lower boundary is exceeded. The results of the simulation still cannot be verified by the results of the implementation.

The demand from the automotive industry for thinner, stronger and more ductile steels led to development of the so called Advanced High Strength Steels (AHSS). These new steel grades are produced at one of the hot dip galvanizing lines, “Dompel Verzinklijn 3” (DVL3), at Tata Steel IJmuiden. The annealing furnace in DVL3 consists out of a number of consecutive sections. One of these sections is the Slow-Cooling Section. The Slow-Cooling Section, a so-called hybrid furnace, is installed with both heating and cooling capabilities to effectively adjust the temperature required by the metallurgical prescription of a specific steel strip product. The required temperature and associated cooling or heating demand can vary from strip to strip, depending on the desired mechanical properties and strip dimensions, which complicates the control of the Slow-Cooling Section. This is further complicated by process variations in production speed and system delays in control instrumentation. It becomes apparent that in order to produce these high quality steel grades, an improved solution to achieve precise temperature control in the slow cooling process has to be found. In this research a Model Predictive Controller (MPC) is developed for the control of the Slow-Cooling Section of DVL3. This advanced control system is based on a mathematical model, developed in the first part of this research, incorporating the heat transfer from and to the steel strips inside the Slow-Cooling Section. The nonlinear furnace model is validated with historical furnace data, showing good agreement for various process conditions. MPC is applied using two different methods. The first method is based on one linear internal model. The second method approximates the nonlinear mathematical model by the use of piecewise affine models. Case studies are performed to test the performance of the new controllers. The performance of the new controllers are compared with the current temperature control of DVL3. A critical discussion is presented based on the results of both Model Predictive Controllers. It is shown that the predictive ability, especially in transient conditions is improved with the newly developed MPC and more accurate temperature control is achieved. Recommendations are given to further improve the control performance.

Dense 3D modeling based on monocular visual data is a powerful process of gaining spatial 3D understanding from 2D observations. The use of visual data to reconstruct such 3D models is still a challenging topic. To obtain the accurate dimensions, additional metadata is required such as a GPS which is not always available. Besides this, dealing with challenging visual situations such as bad-lightening conditions or motion blur remains a difficult subject. Furthermore, since visual data is highly dimensional, most algorithms lack scalability, meaning that they fail to reconstruct 3D models in acceptable time limits and are incapable of handling large data sets. In this thesis, the objective is to mitigate these typical issues whilst preserving the quality of the dense 3D models. To this end, visual-inertial SLAM and MVS techniques are combined to form a dense 3D modeling architecture that is capable of mitigating the typical challenges of classical dense 3D modeling approaches. Besides this, the architecture is extended with additional improvements in the MVS system. These improvements further increase the scalability by abstracting the input data in a highly compact representation by leveraging image segmentation techniques. The result is a novel, visual-inertial dense 3D modeling system. The novel system is tested on benchmark data sets and within a lab setting, where a remote-inspection case-study is performed. The presented system is compared against the industrial and academic state-of-the-art systems. A thorough comparison is made by evaluating the pose accuracy, computation time, and reconstruction quality. It is shown that the presented system improves the state-of-the art systems by a significant margin in terms of computation-time. Furthermore, the presented system is capable of computing 3D models with accurate geometric scale without relying on external metadata, showcasing the effectiveness of the presented system. This work contributes to the lack of research in dense 3D modeling based on visual-inertial SLAM and paves the way for a new direction of efficient MVS algorithms.

An intracranial aneurysm is a bulge in the cerebral vasculature. The rupture of an aneurysm results in a brain bleed. As a consequence, most patients become severely handicapped or may even die. Preventive treatment with endovascular coiling is controversial due to the high risk of complications. These complications are partly caused by the current uncontrolled delivery of coils to the aneurysm. While recent developments in microcatheter design allow for improved positioning, no method has been devised to model and control the tension applied by the coil on the aneurysm wall throughout a coiling procedure. This thesis aims to come up with a method to improve the safety of aneurysm treatment. This thesis presents a new way to dynamically model an endovascular coil and its interaction with a microcatheter and the aneurysm wall throughout a coil deployment procedure. The main advantage of this model is its low computational complexity allowing real-time control computation. A control architecture is presented that enables regulation of the contact force between the coil and the aneurysm wall while obeying the constraints imposed by the equipment and the environment. The presented architecture comprises an augmented energy-shaping controller working in parallel with a constraint preservation controller. This thesis shows that this control architecture asymptotically stabilizes the aneurysm wall tension at the desired value throughout the coil deployment procedure. This work provides a basis for modelling and control in future experimental validations. Therefore, this work is a promising first step in the modelling and control of robotic systems for neurovascular interventions and a step forward in the preventive treatment of intracranial aneurysms.

During the preparation for the Olympic Sailing Competition, held in 2021 in Tokyo, Japan, the Dutch National Sailing Team encountered days with unpredicted wind behaviour. To gain more understanding in the wind patterns occurring, a deep learning based approach is taken. The goal of this research is to find out if unsupervised learning methods can contribute to wind pattern classification. It can then be investigated if the classification can increase understanding in specific wind patterns. The input data for the unsupervised learning model consists of 40 years of reanalysis wind speed data of an area including Japan. To classify the wind patterns, the dimensionality of the input data is reduced using different autoencoders. This reduced dimensional form is then clustered using K-means clustering. The results of the K-means algorithm are compared and the best autoencoder is chosen. The resulting clusters are analyzed for extreme wind patterns, such as typhoons. It is expected that these wind patterns will be clustered together. To check this, the cluster containing typhoon Jebi, the typhoon which caused the highest insurance cost ever in Japan, is analyzed. If this cluster contains typhoons, unsupervised learning is able to provide useful information regarding wind patterns. The best working autoencoder used in this research is the 3D CNN autoencoder. Using the 3D CNN autoencoder, some clusters with specific wind patterns are found. The cluster containing typhoon Jebi consists of 95.8% of typhoons, from which it can be concluded that unsupervised learning is a valid method for wind pattern classification.