This thesis presents the design, development, and experimental validation of a perching drone equipped with an underactuated robotic gripper and tactile sensing for grasping onto structures. Perching extends drone endurance for applications such as long-term monitoring, while tactile sensing enables precise alignment when visual data is unreliable. A control strategy combining position-based control with tactile feedback is implemented using DIGIT tactile sensors for contact-aware adjustments. Two per-pixel inference models convert RGB images into tactile information: a sensitive contact model for binary contact detection and a depth reconstruction model that estimates surface normals, which is then used to determine the contact surface orientation. After outlier filtering, the depth model achieves a mean absolute error of 5.32° in orientation estimation. Experiments demonstrate reliable grasping with up to 12 cm of position error and successful correction of both position and orientation using simulated tactile input. These results highlight the potential of tactile-based strategies for robust aerial manipulation in uncertain environments.

This thesis presents a simulation of a Martian lander using Reinforcement Learning. The objective is to train an agent to land on Mars using the Proximal Policy Optimization (PPO) method. The spacecraft is controlled by control allocation, where the translations and rotations are controlled independently. Also a PD controller is made to control the same lander.

The PD controller is found to be more accurate in comparison with the Reinforcement learning controller. A reinforcement learning thruster model is also made. This spacecraft is controlled by five thrusters and the motions coupled. This lander needs more control effort, and is less accurate in tracking a reference velocity. It is also less accurate with landing on the designated landing spot.

This paper presents the first decentralized method to enable real-world 6-DoF manipulation of a cable-suspended load using a team of Micro-Aerial Vehicles (MAVs). Our method leverages multi-agent reinforcement learning (MARL) to train an outer-loop control policy for each MAV. Unlike state-of-the-art controllers that utilize a centralized scheme, our policy does not require global states, inter-MAV communications, nor neighboring MAV information. Instead, agents communicate implicitly through load pose observations alone, which enables high scalability and flexibility. It also significantly reduces computing costs during inference time, enabling onboard deployment of the policy. In addition, we introduce a new action space design for the MAVs using linear acceleration and body rates. This choice, combined with a robust low-level controller, enables reliable sim-to-real transfer despite significant uncertainties caused by cable tension during dynamic 3D motion. We validate our method in various real-world experiments, including full-pose control under load model uncertainties, showing setpoint tracking performance comparable to the state-of-the-art centralized method. We also demonstrate cooperation amongst agents with heterogeneous control policies, and robustness to the complete in-flight loss of one MAV. Videos of experiments: https://autonomousrobots.nl/paper_websites/aerial-manipulation-marl

Effective scheduling is challenging in Component Maintenance, Repair, and Overhaul (CMRO) operations due to the complexity of dynamically allocating resources across multiple jobs with varying priorities and technical constraints. Current industry practices typically rely on static, manual scheduling, resulting in suboptimal resource allocation and insufficient adaptability to operational disruptions. Most existing studies approach specific job shop problems by incorporating individual features, such as job prioritization or resource constraints, without considering the combined operational complexities of CMRO shops. Therefore, this research presents a scheduling model using the Flexible Job Shop Scheduling Problem (FJSSP), tailored to dynamic CMRO environments. The model uses Mixed Integer Linear Programming (MILP) to simultaneously schedule technicians and machines while accounting for skill requirements, resource constraints, and job prioritization. The approach balances multiple objectives, including tardiness and earliness, to enhance shop performance metrics such as Turnaround Time (TAT) and On Time Delivery (OTD) rates. Outcomes from a case study applied to real-world data from CMRO shops demonstrate significant operational improvements, achieving a reduction in TAT of up to 34% and an improvement in OTD by approximately 23% relative to historical shop performance. Furthermore, the model incorporates schedule robustness measures, minimizing deviations from planned schedules, despite operational uncertainties. Additionally, comparative analysis with a traditional heuristic dispatching rule model confirms the superior performance of the proposed optimization framework. This framework can be broadly applied to improve scheduling efficiency and stability in CMRO shops and similar workshop environments.

Aviation contributes to global warming not only through its $\text{CO}_2$ emissions, but also through contrail formation. The latter may be combatted by taking this effect into account in flight trajectory optimization. Research already shows some possibilities of doing this, but often lacks integration with the operational limits that airlines need to adhere to. In this paper, an approach is described to first optimize flight trajectories for persistent contrail minimization and then show the feasibility of these trajectories as standard flight plans. This is done by using the open source OpenAP.top trajectory optimizer and atmospheric data. An entire day of European flights in and out of Schiphol are analyzed in determining the potential reduction of climate impact for these flights. It was demonstrated that the amount of persistent contrails can be reduced with optimized trajectories and it was shown that these trajectories can, to a large extent, be approximated by waypoints that are included in standard flight plans.

This thesis presents a robust, signal-based mixed-sensitivity H-Infinity controller specifically designed for the flare phase of automatic landings. A critical literature review highlights the shortcomings of conventional gain-scheduled and nonlinear approaches, motivating a multi-objective H-Infinity formulation that treats flare as a stringent mixed-sensitivity problem. Since touchdown occurs in a rapidly changing aerodynamic environment, the controller must deliver tight regulation of vertical speed and pitch attitude while tolerating ground effect, parameter variations and wind disturbances.

A reduced three-state longitudinal model, extracted from a 16-state benchmark and trimmed at 5 m height, captures the most critical seconds before touchdown for controller synthesis. In systune, eleven hard frequency-domain objectives sculpt the principal sensitivity functions and more. Five static feedback gains and a first-order feed-forward filter are tuned simultaneously, achieving 7 dB / 43 deg robust stability margins.

Frequency-domain checks and extensive nonlinear simulations, including a large-scale Monte-Carlo campaign, confirm consistent performance: the probability of exceeding a 0.914 m/s sink-rate is just 0.0029 %, demonstrating precise vertical-speed control across wide operating envelopes. These results show the synthesized controller significantly enhances safety margins by maintaining stability, reducing touchdown dispersion, and fortifying against model and environmental uncertainties. Thus, this research contributes valuable insights toward enhancing the robustness and reliability of future automated landing systems.

The assignment of cargo shipments to available capacity is a complex process. Shipments that vary greatly in weight, volume, and shape must be loaded into Unit Load Devices (ULDs). A slight overestimate of the required number of ULDs can be very costly in terms of delayed cargo. This paper presents a decision-support tool that provides packing strategies for the handling & operations department of a cargo airline. At regular intervals, advice is provided concerning: when to build up a ULD, what items to place in the ULD, and where to place the items within the ULD. As the model considers all flights handled by the warehouse, it is able to optimize operations on a macro-scale rather than an individual flight level. The scheduling of build-up requires accurate information on the availability time of cargo at the warehouse. to deal with uncertainty inherent to such availability, bespoke machine learning prediction models are developed which can provide prediction distributions rather than point estimates. The performance of the model is evaluated through a simulation with real-world data from historic operations of a partner airline. Particularly, the effects of the uncertainty of cargo arrival times on loading performance is investigated. The method to deal with this uncertainty is implementing staff buffer as a contingency. The tool can analyse the cost and benefit of introducing staff buffer to find the optimal amount for each scenario.

This thesis investigates the use of H-infinity Open Loop Shaping (OLS) for designing Thrust Vector Control (TVC) systems during the atmospheric ascent of flexible launch vehicles (LVs). The research has two main goals. The first goal is to assess whether H-infinity OLS can produce rigid body controllers for launch vehicles with comparable or superior robustness, stability, and performance to controllers designed with H-infinity Closed Loop Shaping (CLS), while also reducing design complexity and time. The second goal is to explore an integrated H-infinity OLS approach that tunes both the rigid body controller and the bending filter simultaneously, simplifying the traditional separate design process.
Results demonstrate that H-infinity OLS and H-infinity CLS yield controllers with identical performance, robustness, and stability when applied to design controllers with the same order and structure. However, H-infinity OLS significantly simplifies the design process by avoiding the manipulation of multiple transfer functions, offering higher reproducibility between flight points, and automatically ensuring robustness at both the plant input and output. Although these benefits come at the cost of converting requirements into open loop specifications and setting loop shaping weighting filters to meet the specifications, both challenges have been addressed and simplified in this thesis.
The thesis also extends the rigid body H-infinity OLS design process into an integrated methodology that simultaneously tunes both the rigid body controller and the bending filter, achieving identical results to the traditional separate design approach while reducing overall design time and complexity.
This research demonstrates that H-infinity OLS provides a robust framework for simplifying the design of TVC systems for launch vehicles, without compromising performance and robustness. Additionally, the integrated H-infinity OLS methodology for tuning both the rigid body controller and bending filter effectively streamlines the design process while achieving identical results to the traditional separate approach. These findings validate H-infinity OLS as a viable alternative to H-infinity CLS and introduce a more efficient methodology for designing both the rigid body controller and the bending filter for TVC control systems in launch vehicles.

UAVs have become increasingly popular, finding applications in diverse areas such as military operations, search and rescue, delivery services, wireless communication, and aerial surveillance. A critical aspect of UAV operations is autonomous landing, especially on moving targets like ships, which presents significant challenges due to the dynamic and unpredictable maritime environment. This research explores the potential of reinforcement learning as a strategy for achieving optimal guidance during the autonomous landing process of the VSQP. The study illustrates that the reinforcement learning learning framework can effectively steer the VSQP, ensuring safe and accurate landings on a moving ship, and outperforming the benchmark controller in a more intelligent manner. The proposed approach not only improves landing performance but also extends the operational capability of the VSQP.

This paper presents a mathematical modeling framework designed for air cargo operations of a combination airline, focusing on intercontinental passenger aircraft. This framework optimizes the palletization problem (assigning cargo to Unit Load Devices (ULDs)) and the weight and balance problem (positioning ULDs within the belly of the aircraft). Unlike the traditional approach of solving these problems in sequence, our framework employs a feedback loop between an integrated one-dimensional bin packing with weight and balance and a three-dimensional bin packing. This feedback loop iteratively refines the solution by backpropagating information when items do not fit, reducing the solution space and computation time. The palletization problem is addressed with a three-dimensional bin packing heuristic, while the integrated onedimensional bin packing and weight and balance problem is solved using a mixed integer linear programming formulation. The model employs a lexicographic optimization approach to prioritize maximizing the percent mean aerodynamic chord at zero fuel weight, leading to significant fuel reduction. Additionally, operational objectives aligned with our partner airline’s practices are optimized to reflect actual operations. Scenario analyses validate the model, demonstrating potential fuel savings across various intercontinental routes and aircraft types, with an average fuel saving of 0.53% and reaching up to 1.90%, yielding substantial economic
and environmental benefits.

Neglecting actuator dynamics in nonlinear control and control allocation can lead to performance degradation, especially when considering fast dynamic systems. This thesis provides a novel method to account for actuator dynamics in the control allocation solution, dynamic incremental nonlinear control allocation, or D-INCA. The incremental approach allows for the implementation of a first order discrete-time actuator dynamics model in the quadratic programming (QP) solver. This model is used to find the optimal command inputs in addition to the desired physical actuator deflections, hereby compensating for actuator dynamics delays. Whereas, the baseline incremental nonlinear control allocation (INCA) approach requires pseudo-control hedging of the outer loop reference to increase closed loop stability margins under actuator dynamics delays. To its advantage, D-INCA does not require feedback of higher order output derivatives than INCA and can be used with nonlinear non-control affine systems. Furthermore, with adaptive D-INCA, or AD-INCA, an actuator dynamics parameter estimator is introduced to adapt the actuator model online, minimizing actuator tracking errors after actuator failures. The proposed methods are applied to a fighter aircraft model with an over-actuated innovative control effectors suite and results are compared to the baseline INCA controller.

Climate change poses a serious threat to ecosystems and increases the need for accurate and rigorous monitoring of ecosystems. Current monitoring solutions are often bulky, expensive, and lack critical functionalities such as on-board inference capabilities, robust wireless connections, and a diverse sensor suite. Ecological monitoring projects often suffer from inefficiencies caused by the large time delays between collecting data and analyzing said data, as well as having to spend large amounts of time in the field setting up the sensors manually. This thesis addresses many of these issues by designing a sensor with an extensive sensor suite, robust wireless capabilities and an on-board audio classifier able to perform real-time inference. Furthermore, attention is paid to making the system extendable in the future and allow for potentially integrating the sensors with a drone delivery- and retrieval system. The system tests performed indicate that the system has great potential given more time to tweak some of its identified shortcomings.

This research proposes a novel reconfigurable and force-balanced aerial manipulator design for fast variable payload tasks. Its force-balancing properties allow for fast end-effector movements while minimizing disturbances introduced to the aerial platform. The manipulator is composed of three pantograph legs connecting the end-effector to the drone base. Each pantograph is equipped with two moving counter-masses that provide the balancing properties to the manipulator. The counter masses are moved by fast linear actuators allowing the manipulator to be force-balanced for different payloads. Extensive testing, performing end-effector trajectory tracking tasks, was performed both on a floating base setup and in flight. The results indicate that the manipulator significantly decreased the reaction forces transmitted to the base. Specifically, it achieved a 45% reduction when comparing the unbalanced and balanced configurations, and a 17% reduction when these configurations included a 53 [g] payload. The drone's position-tracking error during flight also improved, with reductions of 19% and 34% for the same two configurations, respectively.

Tiny object detection (TOD) in satellite imagery is critical for applications including pipeline monitoring, where the detection of tiny objects, such as excavators near the pipeline networks, can prevent potential incidents. However, TOD faces challenges due to the limited pixel representation of objects, complications with Intersection over Union-based label assignment, and semantic confusion between visually similar classes. This paper introduces OE-Net, a modular extension of Oriented R-CNN,
specifically designed for TOD in satellite images. OE-Net integrates three novel components: a Dilated Fusion module to enhance the feature extraction, a Balanced Ranking Assigner to improve the anchor matching, and Vector Mapping for more precise classification of visually similar objects. The modular design of OE-Net allows its novel components to be easily integrated into other detection models. Tested on the TinyDOTA dataset, OE-Net sets a new benchmark in TOD performance, achieving a 6.3
percentage points improvement in mean Average Precision (mAP) over the Oriented R-CNN baseline. On a novel excavator detection dataset, named ExcaSat and developed for pipeline monitoring, OE-Net outperforms the Oriented R-CNN baseline by 11.1 percentage points in mAP. Furthermore, OE-Net surpasses state-of-the-art methods on both TinyDOTA and ExcaSat, establishing a new standard in computational efficiency and detection precision for satellite-based monitoring tasks. The code is available at: https://github.com/OE-jimvanoosten/mmrotate-OE.git.

Proximity operations around asteroids are currently at the forefront of space research and industry due to their strategic and scientific relevance. As recent missions increasingly rely on autonomous systems to reduce operational costs, maintaining low-altitude orbits around asteroids without support from the ground represents a key challenge in enabling the transition to a fully autonomous mission architecture.

This research proposes a sliding mode control framework to ensure robustness against navigation uncertainties and gravitational perturbations while accommodating the on/off operating mode of conventional monopropellant thrusters. The designed station-keeping scheme was tested for a range of orbits with varying sizes and orientations around asteroid Eros, demonstrating that the spacecraft can reach altitudes as low as 2 km by relying exclusively on the knowledge of the asteroid’s gravitational parameter. A line-of-sight tracking mode, replacing the conventional two-satellite formation control approach, reduces station-keeping costs and increases coasting time, allowing more opportunities for scientific operations.

We conduct a simulation study of an insect-inspired navigation method that combines visual learning in a small area around a home location with path integration to successfully navigate over distances 8 to 10 times larger than the learning radius, while only requiring 6MB of memory. Our simulations show that the method is capable of learning with on-board data only and handling the noise levels and errors it may encounter in real-world conditions. Furthermore, we investigate the integration of online learning with a structured buffer to mitigate catastrophic forgetting, allowing continuous learning during flight. Additional experiments assess the system's generalization capabilities beyond the learned area, its robustness to small variations in pitch and roll, and the relationship between network size and the learnable area size.

Autonomous drone racing has gained attention for its potential to push the boundaries of drone navigation technologies. While much of the existing research focuses on racing in obstacle-free environments, few studies have addressed the complexities of obstacle-aware racing, and approaches presented in these studies often suffer from overfitting, with learned policies generalizing poorly to new environments. This work addresses the challenge of developing a generalizable obstacle-aware drone racing policy using deep reinforcement learning. We propose applying domain randomization on racing tracks and obstacle configurations before every rollout, combined with parallel experience collection in randomized environments to achieve the goal. The proposed randomization strategy is shown to be effective through simulated experiments where drones reach speeds of up to 70 km/h, racing in unseen cluttered environments. This study serves as a stepping stone toward learning robust policies for obstacle-aware drone racing and general-purpose drone navigation in cluttered environments. Code is available at https://github.com/ErcBunny/IsaacGymEnvs.

Aerial physical interaction opens the door for many operations at height to be automatised using aerial robots. This research presents a novel manipulator design mounted on a traditional quadrotor, which utilises both mechanical and software compliance to perform physical interaction on vertical walls and overhanging surfaces, such as those found under bridges. A centralised impedance control scheme allows direct control of the end-effector pose without needing separate modes for free-flight and contact. A spring-loaded prismatic joint provides passive compliance while doubling as a force-feedback for the impedance controller through measuring the spring displacement. Simulation and flight experiments prove the feasibility and robustness of this approach for exchanging high forces at height, with a total of 44 successful experiments carried out in four sets. An average maximum force of 5.66 N or 19.3\% of the system's weight was achieved over one set of 11 experiments.

Self-supervised deep learning methods have leveraged stereo images for training monocular depth estimation. Although these methods show strong results on outdoor datasets such as KITTI, they do not match performance of supervised methods on indoor environments with camera rotation. Indoor, rotated scenes are common for less constrained applications and pose problems for two reasons: abundance of low texture regions and increased complexity of depth cues for images under rotation. In an effort to extend self-supervised learning to more generalised environments we propose two additions. First, we propose a novel Filled Disparity Loss term that corrects for ambiguity of image reconstruction error loss in textureless regions. Specifically, we interpolate disparity in untextured regions, using the estimated disparity from surrounding textured areas, and use L1 loss to correct the original estimation. Our experiments show that depth estimation is substantially improved on low-texture scenes, without any loss on textured scenes, when compared to Monodepth by Godard et al. Secondly, we show that training with an application's representative rotations, in both pitch and roll, is sufficient to significantly improve performance over the entire range of expected rotation. We demonstrate that depth estimation is successfully generalised as performance is not lost when evaluated on test sets with no camera rotation. Together these developments enable a broader use of self-supervised learning of monocular depth estimation for complex environments.