Courses

@mastersthesis{uuid:6b1796a0-4c41-43b9-ae07-f11ce44f1f83,

title = {Asynchronous Event-Based Visual-Inertial Odometry for Flapping-Wing MAVs},

author = { T. Vennink},

url = {https://resolver.tudelft.nl/uuid:6b1796a0-4c41-43b9-ae07-f11ce44f1f83},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Flapping-Wing Micro Air Vehicles (FWMAVs) operate under severe Size, Weight, and Power (SWaP) constraints, and their high-frequency flapping produces strong oscillatory dynamics that degrade conventional frame-based visual perception. Event cameras are well suited to these conditions, but event-based visual-inertial odometry (EVIO) methods typically rely on short time-window, locally linear motion assumptions that flapping flight breaks: under flapping, event-alignment methods such as plane fitting and contrast maximization recover mostly the high-frequency flapping-induced body rotation rather than the platform's translation. This work presents the first full EVIO pipeline running entirely on-board a small, high-frequency flapping-wing platform. The front-end deliberately avoids short-window alignment and instead exploits the abundance of flapping-induced events: it detects corner features event-by-event, clusters them into compact spatiotemporal features, and tracks these while discarding the fine oscillatory structure. The resulting feature tracks are fused with inertial measurements in a tightly coupled Extended Kalman Filter. The complete pipeline runs in real time on an on-board microcontroller within a sensing-and-compute payload of roughly 21 g. On flight datasets recorded on-board the FWMAV, the pipeline recovers velocity and attitude during flapping flight, and in a closed-loop flight test the on-board position estimate is accurate enough to hold a stationary hover in the horizontal plane. The same event stream that breaks conventional EVIO assumptions is thus sufficient not only for on-board state estimation, but for closing a position-control loop in flight.},

note = {de Croon, G.C.H.E. (graduation committee member); Shyrokau, B. (graduation committee member); de Wagter, C. (graduation committee member); Kober, J. (mentor); Stroobants, S. (mentor); Delft University of Technology (degree granting institution); graduation 24-07-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:94266598-acb5-498f-8316-da3e4a001f3c,

title = {An Active Perception Framework for Assessing Object Pushability with Aerial Robots},

author = { E.K. Cho},

url = {https://resolver.tudelft.nl/uuid:94266598-acb5-498f-8316-da3e4a001f3c},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Aerial robots typically treat surrounding objects as obstacles, even when some could be pushed aside. This thesis presents a vision-based active-perception framework for assessing whether an aerial robot has sufficient force to push an unknown object. The framework includes an uncertain outcome to indicate when visual estimates do not support a definite decision and physical probing would be needed; executing physical probing and pushing is outside the scope of this work. Pushability is modeled relative to the robot's available horizontal force and the object's estimated resistance to sliding. Active multi-view voxel reconstruction estimates object volume, while material classification provides a class associated with predefined density and static-friction coefficient intervals. These estimates form a friction-force interval that, together with a material-confidence check, produces one of three outcomes: pushable, not pushable, or uncertain. In the integrated Unity evaluation, no object was incorrectly classified as pushable. However, only 10 of 22 force-feasibility decisions for objects with accepted material predictions matched the reference outcome, and six of seven reference-ambiguous objects were classified as not pushable rather than uncertain. Counterfactual analysis linked most errors to limitations in the density–friction model, with a plausible contributing explanation being the mismatch between solid-material density assumptions and the effective densities of hollow or composite objects. In Gazebo, two of eight objects received the intended decision, while three could not complete viewpoint acquisition. The framework provides a baseline for visual assessment of aerial pushing force feasibility, while the results highlight the need for more representative physical-property models, explicit treatment of volume uncertainty, and integration of physical probing.},

note = {Popovic, M. (mentor); Hamaza, S. (mentor); Delft University of Technology (degree granting institution); graduation 03-11-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:0b624189-f88b-490e-ad4f-fb90951d5fef,

title = {Energy-Aware Multi-UAV Coordination using Informative Path Planning for Maritime Search and Rescue: Exploiting Flight-Regime Energy Dynamics},

author = { S.D. Bhawan},

url = {https://resolver.tudelft.nl/uuid:0b624189-f88b-490e-ad4f-fb90951d5fef},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Autonomous maritime search and rescue faces critical challenges due to limited endurance and dynamic victim drift. This thesis presents a centralized multi-UAV coordination framework based on Energy-Aware Informative Path Planning. The aim is to maximize mission endurance by exploiting the distinct energy dynamics of different flight regimes, such as the efficiency of fixed-wing cruise versus the high cost of hovering. Locally, a Receding Horizon Planner optimizes the trade-off between entropy reduction and energy consumption. The framework provides global coordination by dynamically switching agents between high-altitude exploration and low-altitude target tracking while accounting for the effects of ocean drift. Monte Carlo simulations using the empirical power model of the Variable Skew Quad Plane evaluate the swarm size required to achieve 80% global entropy reduction and successfully confirm multiple victims within a strict 10-minute timespan. The evaluation reveals a fundamental trade-off between spatial clearing speed and environmental robustness—defined as the system's ability to maintain consistent search performance despite adverse wind disturbances. Although boustrophedon coverage is efficient in optimal conditions, adverse crosswinds cause accelerated energy drain, potentially leading to early mission failure. By dynamically surfing wind gradients, the proposed method extends mission endurance while maintaining sufficient coverage capabilities in these severe crosswind scenarios. Finally, the communication and coordination logic of the framework was validated through real-world flight tests with a swarm of Parrot Bebop 2s.},

note = {Popovic, M. (mentor); Remes, B.D.W. (mentor); Delft University of Technology (degree granting institution); graduation 23-04-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:24578072-a7e5-46ba-8baa-79896cb09f3c,

title = {Hardware-in-the-loop simulation for Earth-based testing of On-Orbit Servicing using Soft Continuum Robotics},

author = { A.H.H. Dourleijn},

url = {https://resolver.tudelft.nl/uuid:24578072-a7e5-46ba-8baa-79896cb09f3c},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Soft continuum robots offer several advantages compared to rigid manipulators when used for on-orbit servicing (OOS) due to their inherent compliance and infinite degrees of freedom. However, these advantages also mean that their behaviour is heavily influenced by the presence of gravity, making representative testing on earth impossible without gravity compensation. This report presents a novel hardware-in-the-loop (HIL) testing setup that allows for representative, terrestrial testing of rigid-soft spacecraft dynamics. Dynamical models of a hybrid chaser spacecraft, featuring soft continuum manipulators (CSM) mounted on a rigid base, and a rigid target body are considered. The hybrid chaser is modelled using the geometrical variable strain approach using a single set of generalised coordinates to represent both the rigid body movement as well as CSM deformations allowing dynamical coupling between the base and the CSMs. A contact model is used to model interactions of the CSMs with the target body. In the HIL setup the relative dynamics between the chaser and the target are reproduced using a rigid manipulator on which the target body is mounted. External forces on the target body are recorded using a force-torque sensor mounted between the body and the manipulator. The setup is tested using two OOS scenarios: successful capture and single CSM interaction. Experiments show that positional errors between the simulated and reproduced bodies on the testbed lie are in the magnitude of 1 cm, both during approach and capture phases. Despite concurrent force measurements from both the FTS and the contact model, the coupled simulation remains numerically stable, the system energy remains bounded and CSM state in simulation as on the testbed remain closely matched. These results demonstrate that the proposed HIL framework can accurately reproduce the coupled dynamics of rigid-soft spacecraft interactions, providing a practical platform for terrestrial validation of continuum manipulator concepts for on-orbit servicing.},

note = {Della Santina, Cosimo (mentor); Popovic, M. (mentor); Delft University of Technology (degree granting institution); graduation 27-08-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:2f3427d8-7779-41ab-8915-ca38e58edb1f,

title = {Continuous-Time Modelling of Pixel Attributes: The Development of a Continuous Frame Interpolation Method and a Fast, Real-Time Event Camera Simulator},

author = { R.J.N. Zwikker},

url = {https://resolver.tudelft.nl/uuid:2f3427d8-7779-41ab-8915-ca38e58edb1f},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {There is an increasing interest in Event-based cameras for computer vision purposes. These cameras provide benefits such as high temporal resolution and high dynamic range by capturing the change in brightness instead of the absolute brightness signal. In order to train learning-based behaviours with event-based cameras, it is essential to be able to simulate the output of these cameras in a virtual world. Unlike their traditional frame-based counterparts, there are no real-time event-based camera rendering engines yet. All existing methods of generating events from virtual scenes impose a significant computational burden, simulating events by generating frames at a very high temporal resolution and converting these frames to events. This thesis proposes a novel method of event generation, by continuously interpolating between frames at a low temporal resolution, using auxiliary data from the rendering engine, such as depth and optical flow. In this method, the backwards pixel sampling problem is solved using a triangle grid created between all pixels, that is used to model the continuous response of each pixel between rendered frames. Other than previous methods that used VFI to temporally upsample LFR frames for event generation, this method does not create intermediate frames, instead simulating events directly from the continuous model. This method is implemented on the CPU, where it is shown to correctly interpolate the signal between frames. It is also implemented in Unreal Engine, where it is used to generate events in near real-time.},

note = {de Croon, G.C.H.E. (mentor); Wu, Y. (mentor); Delft University of Technology (degree granting institution); graduation 31-03-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:4202cea2-f25f-439d-aa15-cace7cca77cb,

title = {Incremental Nonlinear Dynamic Inversion Control of a Launch Vehicle Testbed},

author = { A.R.S. Bracke},

url = {https://resolver.tudelft.nl/uuid:4202cea2-f25f-439d-aa15-cace7cca77cb},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Launch vehicles traditionally rely on gain-scheduled PID controllers, which are time-consuming to develop, and require extensive modelling. Incremental Nonlinear Dynamic Inversion (INDI) may serve as an alternative to reduce model dependency, and streamline launch vehicle control system design and tuning. This work applies a cascaded INDI controller, with an inner attitude loop and an outer position loop, to the model of a small-scale launch vehicle testbed. Such testbeds are used in the context of Guidance, Navigation & Control (GNC) to develop and test algorithms on a physical system in a cost-effective and low-risk fashion. Wind disturbance, propellant sloshing, and tail-wags-dog, i.e. the effect of the engine’s motion on the rest of the vehicle, are considered in the model. The results show that INDI provides a simple way to develop a nonlinear control law that maintains robust position reference tracking in the presence of disturbance and model uncertainty. Finally, it is determined that a small-scale testbed is not suitable for studying the impact of sloshing on the closed loop control systems of launch vehicles, since at this scale the sloshing mass is too small and the natural frequency of the sloshing dynamics is too high.},

note = {Smeur, E.J.J. (mentor); van Kampen, E. (mentor); de Visser, C.C. (graduation committee member); Mooij, E. (mentor); Acquatella, P. (mentor); Delft University of Technology (degree granting institution); graduation 27-03-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:5640fa42-1c33-4639-838f-5fda5b4cb37f,

title = {Full Actuator Nonlinear Dynamic Inversion for Enhanced Hybrid UAV Control},

author = { J.P.G. Dubois},

url = {https://resolver.tudelft.nl/uuid:5640fa42-1c33-4639-838f-5fda5b4cb37f},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Expanding the operational capabilities of Micro Air Vehicles (MAVs) hinges on control systems that manage highly nonlinear dynamics across broad flight envelopes. Incremental Nonlinear Dynamic Inversion (INDI) is popular for its simplicity and modest modeling needs, but its assumption of infinitely fast actuators and neglect of state-dependent effects limit performance when actuators have slow or heterogeneous dynamics or when aerodynamic effects are significant. Actuator Nonlinear Dynamic Inversion (ANDI) overcomes these limitations by explicitly incorporating state-dependent dynamics and finite actuator bandwidth into the control law, enabling improved tracking performance across diverse actuator configurations. This work implements the full ANDI stabilization controller on the Cyclone, a hybrid MAV tail-sitter, using cascaded complementary filtering for state estimation. Simulation and flight experiments validate the approach and assess whether this compensation yields practical performance gains, establishing ANDI as a viable, generic control solution for MAVs Code is available at: https://github.com/tudelft/paparazzi/tree/feat_stabilization_andi_controller.},

note = {van Kampen, E. (graduation committee member); Smeur, E.J.J. (mentor); Ntouros, E. (mentor); Theodoulis, Spilios (graduation committee member); Delft University of Technology (degree granting institution); graduation 11-02-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:5e8691b1-cf90-4735-90f2-9fef88af0b67,

title = {Anomaly Detection for Cluttered, Multi-Instance E-grocery Stock Containers},

author = { D.J. Jagt},

url = {https://resolver.tudelft.nl/uuid:5e8691b1-cf90-4735-90f2-9fef88af0b67},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Automated quality inspection of cluttered, multi-instance e-grocery stock containers is challenging because packaging clutter, product boundaries, container wear, and loose debris generate local anomaly responses that do not correspond to genuine quality issues. This thesis develops a modular reference-based inspection pipeline in which normality for each stock keeping unit (SKU) is modelled through automatically constructed and sampled reference sets. Query images are scored by patch-level nearest-neighbour matching of DINOv2 features against a reference memory bank, following the AnomalyDINO paradigm. Edge downweighting and crop-level debris filtering suppress boundary- and debris-driven false positives, while an optical character recognition (OCR)-based branch additionally addresses visually subtle wrong-SKU substitutions. Evaluated on 2,996 normal and 1,657 issue images across 20 SKU classes and five issue families, the full pipeline increases precision from 0.511 to 0.834 and reduces false positives from 1,414 to 201 relative to the core detector (F1-score: 0.705). This improvement comes at a recall cost, with recall decreasing from 0.893 to 0.610. The full pipeline achieves a 99th-percentile end-to-end latency of 1.4 s, within the 10 s operational constraint. The results demonstrate that robust inspection in this setting requires treating false-positive mitigation as a core system requirement rather than relying on raw anomaly sensitivity alone.},

note = {Popovic, M. (mentor); de Wagter, C. (graduation committee member); van Gemert, J.C. (graduation committee member); Schultinga, Kevin (mentor); Delft University of Technology (degree granting institution); graduation 28-05-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:7e0c7f9d-f418-486a-93d3-9dadf33a6614,

title = {Safety Cages for Reliable Machine Learning: A framework for exoplanet spectroscopy in the Ariel mission, extended with error-aware model specialisation},

author = { N.I. Grens},

url = {https://resolver.tudelft.nl/uuid:7e0c7f9d-f418-486a-93d3-9dadf33a6614},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Ensuring the reliability of machine learning models in safety-critical space missions remains a significant challenge, especially when ground-truth data is unavailable for real-time validation. While machine learning can augment pipelines by extracting transmission spectra from complex light curves, vulnerabilities to instrument anomalies and domain shifts introduce risks. This study evaluates a modular safety cage architecture operating as a parallel monitoring layer to assess prediction validity without modifying the underlying estimator. By monitoring runtime indicators, including uncertainty quantification, out-of-domain detection, and influence functions, the framework constrains the operational domain to a verified region. The results demonstrate that model failure is multifaceted, requiring indicator fusion strategies, and that applying safety-driven rejection allows for a small reduction in data coverage leading to a significant reduction in prediction error. In addition, the framework is extended with an error-aware model specialisation pipeline that partitions the parameter space to deploy specialised local experts.},

note = {de Croon, G.C.H.E. (mentor); F. Simões, Luís (mentor); Cazaux, S.M. (graduation committee member); Meijer, Vincent (graduation committee member); Delft University of Technology (degree granting institution); graduation 30-06-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:995273bf-d7b7-403d-90ff-546634a673d8,

title = {Tactile Object Classification Using a Biomimetic Whisker Sensor Array and Spiking Neural Networks},

author = { T.J. van der Weij},

url = {https://resolver.tudelft.nl/uuid:995273bf-d7b7-403d-90ff-546634a673d8},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Robots operating in confined or visually degraded environments require sensing modalities that provide reliable close-range feedback. Biologically inspired whisker sensors address this need: passive and lightweight, they encode contact geometry through time-varying signals. Spiking neural networks (SNNs) form a natural processing counterpart — also biologically motivated, they perform sparse, event-driven computation that mirrors neural processing and promises energy-efficient inference on neuromorphic hardware. Despite this complementary fit, training data for neuromorphic tactile sensing remains scarce, and fully spiking architectures have not been evaluated on whisker-based classification, leaving their practical viability an open question. To address this gap, we collected two tactile datasets using a 10-whisker MEMS-based sensor array swept across three object geometries under varying contact conditions. A compact fully spiking network with 901 parameters achieved 100% test accuracy, outperforming larger spiking architectures; classification was reliable from as few as 20 timesteps of contact, and raw pressure signals alone were sufficient. Generalization experiments showed that sweep speed had minimal effect on performance, while indentation depth and sweep direction introduced larger domain shifts. Compared to a structurally equivalent non-spiking baseline, the spiking model matched classification accuracy with an estimated 9.5% reduction in overall inference energy under analog input and a potential 95% reduction under spike-encoded input. These results demonstrate the feasibility of a fully bio-inspired pipeline — from biomimetic whisker sensing to spiking neural inference — for tactile object classification.},

note = {Hamaza, S. (mentor); Tömen, N. (mentor); Ferranti, L. (mentor); Delft University of Technology (degree granting institution); graduation 22-04-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:a4248cc6-a95a-481a-9065-965987da8d83,

title = {Local Path Planning and Obstacle Avoidance for an Omnicopter: A 6D-DWA Algorithm for Real-time Omnidirectional Navigation},

author = { M.K. Heliński},

url = {https://resolver.tudelft.nl/uuid:a4248cc6-a95a-481a-9065-965987da8d83},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {As omnidirectional UAVs become increasingly prevalent, computationally efficient path planning algorithms are required to facilitate their autonomous traversal of dynamic environments. This thesis presents a real-time local path planning algorithm for an omnicopter, adapted from the Dynamic Window Approach into a 6D-DWA implementation. To ensure computational efficiency, the local path planner incorporates environment voxelisation, a 10-sphere omnicopter structure approximation, and adaptive velocity sampling. The avoidance of dynamic obstacles, detected via depth camera, is facilitated through context-aware 6D-DWA weight adjustments and a separate integrated fast-evasion strategy. Simulation testing in Gazebo demonstrates that the 6D-DWA is a feasible local planning solution, consistently performing below the 0.2 s real-time threshold. Future research into algorithm hybridisation may further mitigate the system's limited foresight and enhance the replanning capabilities.},

note = {Pavel, M.D. (graduation committee member); de Wagter, C. (graduation committee member); Popovic, M. (mentor); Theodoulis, Spilios (mentor); Delft University of Technology (degree granting institution); graduation 05-03-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:bd77521e-91c5-46b4-8349-c847ce8bac78,

title = {Autonomous fixed-wing interception using reinforcement learning},

author = { D.F. van Hamersveld},

url = {https://resolver.tudelft.nl/uuid:bd77521e-91c5-46b4-8349-c847ce8bac78},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Medium altitude, long endurance unmanned aerial systems such as Shahed-type strike drones are an unanswered threat in modern conflicts. Conventional air defense systems are too expensive to provide a feasible large-scale solution. Quadcopter interceptors are a cheap, battle-proven alternative, but their extremely limited flight time restricts them to low altitude targets at short range, whereas fixed-wing platforms have the endurance to reach higher altitude targets from longer distances. However, such interceptors are still operated manually. Recent developments in learning-based controllers show potential for automation. In this work, a fixed-wing interceptor is presented that is controlled by a neural network that maps filtered monocular vision observations directly to actuator commands. The policy was trained in a custom Python environment using reinforcement learning and proximal policy optimization. The policy was then deployed on a fixed-wing platform and tested in real interception scenarios. As a baseline, two classical guidance controllers were also developed and tested in simulation. The policy is able to intercept targets both in simulation and in real flight. In simulation, the policy outperforms the classical controllers under most conditions. The interceptor was ultimately successfully validated in real flight by intercepting maneuvering fixed-wing targets. This novel system shows the potential of artificial intelligence for autonomous control of fixed-wing interceptor drones.},

note = {Smeur, E.J.J. (mentor); Kriachko, I.T. (mentor); Delft University of Technology (degree granting institution); graduation 04-09-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:cb78a143-7cf2-4593-9afa-ca2ca900b1b0,

title = {Active Exploration for VLM-Guided Anomaly Inspection using a UAV},

author = { T.L. van der Wal},

url = {https://resolver.tudelft.nl/uuid:cb78a143-7cf2-4593-9afa-ca2ca900b1b0},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Autonomous exploration by drones in unknown environments has traditionally focused on maximizing spatial coverage without semantic understanding. This thesis presents a framework that integrates vision-language models (VLMs) with adaptive path planning to enable anomaly-aware exploration and inspection. The system employs a three-phase approach: frontier-based exploration, continuous VLM-based anomaly detection, and inspection of detected anomalies. Comparative experiments demonstrated that YOLO+CLIP with negative embeddings achieved the highest F1 score of 0.7218 on the SegmentifyMeIfYouCan benchmark. Experiments showed that dedicated inspection yielded improvements over solely exploration observations. However, system-level evaluation across nine experimental runs revealed that the inspection phase took up most of the mission time (85.9% average), with varying anomaly detection consistency across anomaly instances. False positive analysis identified VLM error as the primary limitation (52% of false positives), followed by simulation artifacts (37%) and semantic ambiguity (11%). The framework successfully demonstrated the feasibility of coupling VLM-based anomaly detection with adaptive planning, though precision limitations and large inspection inefficiencies show opportunities for future work.},

note = {Smeur, E.J.J. (graduation committee member); Popovic, M. (mentor); Blum, Hermann (mentor); Della Santina, C. (graduation committee member); Delft University of Technology (degree granting institution); graduation 19-03-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:ecd458f0-6a6b-4295-a998-619bceaeb00c,

title = {Towards Decoding Motion: How Depth and Optical Flow Inform Neural Ego-Motion Estimation},

author = { Q. Missinne},

url = {https://resolver.tudelft.nl/uuid:ecd458f0-6a6b-4295-a998-619bceaeb00c},

year  = {2026},

date = {2026-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Deep neural networks are increasingly used for ego-motion estimation. Often, in self-supervised ego-motion networks, it is decoded from a depth network and until now has not been decoded from an optical flow network. This is surprising given the tight relationship between optical flow and ego-motion. While both representations are widely used in learning-based approaches, the extent to which their latent space encodes motion information remains poorly understood. This paper presents a controlled analysis of how well depth-based and optical flow-based neural networks encode ego-motion. Using supervised depth, flow and pose networks trained on TartanAirV2, we probe motion information by attaching identical minimal pose decoders to frozen encoders. Representational space analysis through Centered Kernel Analysis (CKA) and feature space analysis through Principal Component Analysis (PCA) are used to examine how motion information is structured across network hierarchies. This work shows that optical flow representations encode ego-motion information more explicitly, in a lower dimensional, linearly accessible structure as opposed to depth representations, which exhibit weak alignment with pose. These findings suggest that in a self-supervised setting, ego-motion estimation can best be decoded from an optical flow network as opposed to a depth network.},

note = {Kooij, J.F.P. (mentor); de Croon, G.C.H.E. (mentor); Zaffar, M. (mentor); Bahnam, S.A. (mentor); Delft University of Technology (degree granting institution); graduation 28-01-2026},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:0ad44d53-6707-4ed5-aa25-07a4e3659f16,

title = {Pedestrian Dead Reckoning using Data-Driven and Physics-Informed Machine Learning},

author = { L. Ligthart},

url = {https://resolver.tudelft.nl/uuid:0ad44d53-6707-4ed5-aa25-07a4e3659f16},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Many applications such as indoor navigation, search and rescue operations, and surveying, require accurate localization in challenging environments. In environments where absolute positioning (using for example GNSS receivers) fails due to dense obstruction of the required signals, positioning algorithms rely solely on dead reckoning, often using measurements from an Inertial Measurement Unit (IMU). This dead reckoning process suffers from integration drift due to the accumulation of errors in the sensor measurements. Pedestrian Dead-Reckoning (PDR) algorithms can reduce this integration drift in cases where the IMU is held by a walking person, by leveraging patterns from the periodic walking motion. This thesis investigates a state-of-the-art PDR algorithm from Liu et al. that combines an Extended Kalman Filter (EKF) with a velocity-predicting neural network that corrects the filter in the measurement update to mitigate integration drift. The focus is on finding if its performance can be improved by adapting the neural network in the algorithm. First, adaptations of the network’s parameters have been experimented with to investigate their effect on the algorithm’s accuracy and search for a bottleneck that limits it. This bottleneck was found to be a bias in the predicted velocity by the network, as this violates the EKF assumption that the error in the measurement update is distributed with zero mean. The second part investigates how using ideas from the Physics-Informed Neural Network (PINN) as an alternative to the data-driven neural network in the PDR algorithm affects its performance. Four network architectures have been trained using a loss function that includes a penalty for errors in the physics of the predicted velocities of the system. Training these PINN-inspired networks required a much longer training time. The results show that using this physics loss does not show significant improvements in the accuracy of the PDR algorithm, as the bias in the velocity prediction is not addressed by the physics loss. This thesis concludes that the main limitation of the neural network in the PDR algorithm by Liu et al. is a bias in the predicted velocity, and that a PINN is unable to provide significant improvements in the accuracy of the algorithm, whilst costing much more computational resources to train. It is therefore not recommended to use a PINN in this PDR algorithm, and the optimal configuration of the network was found to be a slightly adapted version of the network used by Liu et al., using IMU and device tilt data sampled at 50 Hz as network input. Future research could focus on understanding the origin of the bias in velocity predictions and to mitigate it once its source is known.},

note = {Kok, M. (mentor); van Bavel, H.R. (mentor); Kraaij, R.C. (graduation committee member); Popovic, M. (graduation committee member); Delft University of Technology (degree granting institution); graduation 10-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:0e1d8dae-4592-4209-be5d-6aac10057491,

title = {Decentralized Aerial Manipulation of a Cable-Suspended Load using Multi-Agent Reinforcement Learning},

author = { J. Zeng},

url = {https://resolver.tudelft.nl/uuid:0e1d8dae-4592-4209-be5d-6aac10057491},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {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},

note = {Sun, S. (mentor); Matoses Gimenez, A. (mentor); Vinitsky, Eugene (mentor); Alonso-Mora, Javier (mentor); Pek, C. (graduation committee member); Smeur, E.J.J. (graduation committee member); Delft University of Technology (degree granting institution); graduation 17-07-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:338ecfab-95f5-4968-87b5-2bd5bdb62917,

title = {Tether-Inertial Localization for Drones on Mars},

author = { D.I. van Loon},

url = {https://resolver.tudelft.nl/uuid:338ecfab-95f5-4968-87b5-2bd5bdb62917},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Recent developments in planetary exploration have shown the potential of Unmanned Aerial Vehicles (UAVs), such as the Ingenuity helicopter that provided valuable mapping data. However, limited payload capabilities constrain the flight times and compute available for localization which restrict their applicability. By providing a tethered connection, issues such as battery and computational constraints are offloaded to the base rover. At the same time, the cable can be exploited for non-drifting localization. This work presents a novel Tether-Inertial Localization approach that uses tether length, and angle measurements to estimate the UAV pose relative to its base. The method combines a computationally efficient analytical catenary model with Gaussian Process (GP) residual error compensation. This accounts for systematic sensor inaccuracies and model limitations under dynamic conditions. Experimental validation across circular, triangular, and figure-eight trajectories with tether lengths up to 4.5 m spanning 41 minutes of flight time with only the tether based estimate as position feedback, demonstrates estimation errors of 0.073 m average RMSE for the analytical model alone and 0.049 m average RMSE with GP-enhancement. With this, we provide an alternative localization method to conventional vision or GNSS-based methods, providing similar accuracies for Tethered UAVs.},

note = {Bredenbeck, A. (mentor); Puck, L. (mentor); Prendergast, J.M. (graduation committee member); Hamaza, S. (graduation committee member); Delft University of Technology (degree granting institution); graduation 24-11-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:e8b6c7fa-a634-45c3-b1f3-2e86a25b88c4,

title = {Autonomous Perching Drones: Design, Perception and Control},

author = { G.M. Strunck},

url = {https://resolver.tudelft.nl/uuid:e8b6c7fa-a634-45c3-b1f3-2e86a25b88c4},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Autonomous perching enables micro aerial vehicles (MAVs) to act as quiet, non-intrusive sensing nodes for ecological monitoring and habitat assessment. We present a 1.2[kg] quadrotor designed for selective branch perching in natural environments, aiming to extend operational endurance and enable elevated observation in dense vegetation. We achieve autonomous quadrotor perching by addressing challenges in perception, planning, mechanical design and control. Our system, equipped with a stereo vision pipeline and a lightweight top-mounted actuated gripper arm, demonstrates fully autonomous perching across a wide range of branch and stem orientations - from horizontal to vertical - even in cluttered, near-natural forest scenes. In controlled laboratory trials, the MAV performs a 360° scan to identify candidate perches, verifies them with depth-based checks, executes a bottom-up, orthogonal approach, and supports repeated take-off and re-perching at different branches. This is the first time the complete pipeline of design, perception, planning, and control is addressed for perching drones. With that, our work lays the foundation for future deployment in ecological monitoring and restoration missions, where minimally invasive aerial observation could benefit remote or sensitive ecosystems.},

note = {Hamaza, S. (graduation committee member); Alonso-Mora, Javier (graduation committee member); Popovic, M. (graduation committee member); Delft University of Technology (degree granting institution); graduation 29-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:0144c330-9b16-4283-9b38-89c1e94a07f4,

title = {Semi-Reactive Collision Avoidance in Unknown Environments},

author = { S. van den Bergh},

url = {https://resolver.tudelft.nl/uuid:0144c330-9b16-4283-9b38-89c1e94a07f4},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Unmanned Aerial Vehicles (UAVs) play a crucial role in various applications, including disaster response, infrastructure inspection, and search-and-rescue missions. To maximise their effectiveness, UAVs must achieve a high level of autonomy, particularly when navigating cluttered environments. This requires real-time collision avoidance algorithms that efficiently utilise on-board computational resources. Existing approaches either maintain explicit map representations, which provide memory of past observations but require significant computation, or directly process depth images, which are computationally efficient but restrict flight paths to the sensor’s field of view (FOV). This thesis presents a novel, lightweight collision avoidance algorithm that operates directly on depth images while retaining memory of past observations. By aggregating previous depth data, the algorithm enables UAVs to plan trajectories beyond the FOV of onboard sensors without the computational overhead of explicit map maintenance. Experimental results demonstrate that the proposed method outperforms existing state-of-the-art methods in terms of computational efficiency, flight speed, energy cost, and path length until collision.},

note = {Popovic, M. (mentor); de Croon, G.C.H.E. (graduation committee member); Guo, J. (graduation committee member); Delft University of Technology (degree granting institution); graduation 01-04-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:0b8dd715-4cb8-4ccc-ab4c-222d01db9d06,

title = {Tactile Aerial Grasping via High-Resolution Touch on Drones},

author = { M.A. Elahi},

url = {https://resolver.tudelft.nl/uuid:0b8dd715-4cb8-4ccc-ab4c-222d01db9d06},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {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.},

note = {Hamaza, S. (mentor); Smeur, E.J.J. (graduation committee member); Bombelli, A. (graduation committee member); Bredenbeck, A. (mentor); Brummelhuis, M.B.J. (mentor); Delft University of Technology (degree granting institution); graduation 29-08-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:18156042-24d1-4bc8-8c6b-d5ea71f978cc,

title = {SkyDreamer: Learning End-to-End Vision-Based Drone Racing with Interpretable Model-Based Reinforcement Learning},

author = { A.E.C. Verraest},

url = {https://resolver.tudelft.nl/uuid:18156042-24d1-4bc8-8c6b-d5ea71f978cc},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

note = {de Wagter, C. (mentor); Ferede, R. (mentor); de Croon, G.C.H.E. (mentor); Bahnam, S.A. (mentor); Delft University of Technology (degree granting institution); graduation 23-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:1fb03e01-f60b-4129-ad36-a3d258a2d14d,

title = {Optical Flow Estimation using SPECK™ Neuromorphic Hardware},

author = { M. Singh},

url = {https://resolver.tudelft.nl/uuid:1fb03e01-f60b-4129-ad36-a3d258a2d14d},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Neuromorphic hardware and spiking neural networks (SNNs) offer a bio-inspired path to low-latency, energy-efficient computation by emulating the brain’s asynchronous spike-based processing, particularly attractive for real-time optical flow estimation on resource-constrained micro aerial vehicles (MAVs). We leverage a SynSense SPECK™ system-on-chip, which integrates a Dynamic Vision Sensor (DVS) with a neuromorphic processor, to realize live, onboard flow-based attitude and thrust control from dense event-based optical flow. Our inference architecture combines spiking and analog layers in a hybrid SNN-ANN framework, enabling the use of SPECK™ for regression task in a closed-loop drone control, an application not previously demonstrated on the chip. Despite the chip’s compact form factor, the system produces dense flow in real time and achieves stable indoor hover using flow-based control. The hybrid pipeline runs ~2× faster than an ANN-only baseline at identical power. These results highlight the promise of neuromorphic sensing and processing for ultra-efficient, autonomous flight in real-world scenarios.},

note = {de Croon, G.C.H.E. (mentor); Ou, D. (mentor); Kober, J. (graduation committee member); de Wagter, C. (graduation committee member); Delft University of Technology (degree granting institution); graduation 29-08-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:22d13eff-4198-4d4b-927a-01edaddd990a,

title = {Beyond Proportional Navigation: Deep Reinforcement Learning for Robust Drone Interception},

author = { M.L. Capelle},

url = {https://resolver.tudelft.nl/uuid:22d13eff-4198-4d4b-927a-01edaddd990a},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {The rapid increase in drone threats has created an urgent need for practical counter-drone systems for safety and defence purposes. Interceptor drones represent a promising and cost-effective solution for removing unwanted aerial vehicles. However, classical guidance laws such as Proportional Navigation (PN) are not well adapted to highly agile targets under realistic perception constraints. In parallel, deep reinforcement learning (DRL) has demonstrated exceptional performance in high-speed quadcopter control tasks, motivating its consideration for interception applications. A major challenge for such learning-based controllers lies in sim-to-real transfer. This work addresses this challenge through extensive domain randomization and explicit modelling of realistic perception constraints, including sensor noise, limited field of view, sensing ranges, and sensor update rates. A framework is proposed that trains Single Rotor Thrust (SRT) DRL policies for quadcopters and evaluates them against a classical PN implementation within a physics-based simulator. The resulting controllers generalise across platform variations and outperform classical guidance under modelled perception limitations, demonstrating improved robustness and transferability. Github Repository of the code used for the thesis: https://github.com/maximecapelle/quadintercept_drl},

note = {Ferranti, L. (mentor); Smeur, E.J.J. (mentor); Alonso-Mora, Javier (graduation committee member); Ferrari, R. (graduation committee member); Delft University of Technology (degree granting institution); graduation 05-11-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:28823cd3-e24b-401e-b328-a10fefb7af78,

title = {Reinforcement Learning-Based Guidance and Control for Aerial-to-Aerial Pest Interception},

author = { M.S. Broekers},

url = {https://resolver.tudelft.nl/uuid:28823cd3-e24b-401e-b328-a10fefb7af78},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {PATS-X is a greenhouse pest suppression system that uses a depth camera with an autonomous micro air vehicle (MAV) to detect, track, and physically intercept flying insects. This study targets guidance and control for reliable aerial-to-aerial interception. Reinforcement learning (RL) is used to learn policies from insect flight recordings. We evaluate control policies at increasing levels of abstraction: direct motor commands, collective thrust and body rates (CTBR), and acceleration. In simulation, lower abstraction levels yield better interception performance; moving from acceleration to motor command reduces the median time to first interception by about 41%. A systematic variation of the observation space reveals that the most effective observations are body frame relative position and velocity, and short temporal histories add no benefit beyond noise filtering. Compared with a state-of-the-art classical benchmark, Fast Response Proportional Navigation (FRPN), the best motor level RL policy in simulation achieves a median first interception time of 0.85 [0.76–1.07]s with 99.1% interception rate, compared with FRPN at 1.90 [1.04–2.80]s and 95.6%. To address the reality gap, we compare how well the different control abstractions transfer to hardware. CTBR policies deploy on hardware with the least performance loss relative to simulation. Motor-level policies also transfer when trained with modest domain randomization (DR) plus an action-difference penalty that limits command jitter and thermal load. Acceleration-level policies did not transfer. In a PATS-X proof of concept, an RL controller deployed on the actual system reached a 95.6% interception rate of virtual moths versus 80.0% for the existing controller. Moreover, the RL controller shortened time-to-first-interception by 0.70s, indicating the potential of RL-based guidance for the PATS-X system.},

note = {de Croon, G.C.H.E. (mentor); Vos, R.W. (mentor); Yedutenko, M. (mentor); de Wagter, C. (graduation committee member); Bombelli, A. (graduation committee member); Delft University of Technology (degree granting institution); graduation 11-12-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:29735fe1-7fe0-4d28-83bd-76c4fa2569ce,

title = {Large-Scale Priority-Aware Exploration with UAV Swarms},

author = { D. Borstlap},

url = {https://resolver.tudelft.nl/uuid:29735fe1-7fe0-4d28-83bd-76c4fa2569ce},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Autonomous UAV swarms have great potential for applications such as search‐and‐rescue, wildfire monitoring, and environmental mapping, but rapid and reliable coverage of large, obstacle‐filled areas remains challenging. In this paper, we ask: How can decentralised multi-UAV systems be optimised for large-scale priority-aware exploration? To answer this, we present the Fast LIDAR-based Autonomous Multi-UAV Explorer (FLAME), a hierarchical framework for efficient swarm exploration with long‐range, 360° LIDAR sensors. Locally, a dual‐mode exploration planner alternates between high‐speed straight‐line travel and frontier-cost‐based exploration. Globally, each UAV’s mission is optimised to visit user‐defined high‐priority regions early while minimising total mission time, enabling decentralised coordination with minimal duplicate exploration. FLAME also integrates battery‐constrained multi‐mission planning, dividing large tasks into multiple battery‐feasible missions with timely returns to the ground station for recharging or battery swaps. In simulation, FLAME achieves 36% lower mission time compared to state-of-the-art. A high-fidelity simulation confirms the framework’s ability to effectively prioritise user-assigned regions while adhering to operational constraints.},

note = {Popovic, M. (mentor); Alonso-Mora, Javier (mentor); Delft University of Technology (degree granting institution); graduation 03-05-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:4fa402ad-8869-42a3-898a-b87d2b6aea65,

title = {Adaptive Reinforcement Learning-Based Control for High-Speed Autonomous Drone Racing},

author = { S.R. Kethiri},

url = {https://resolver.tudelft.nl/uuid:4fa402ad-8869-42a3-898a-b87d2b6aea65},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {This work investigates adaptive switching mechanisms for reinforcement learning (RL) controllers in high-speed autonomous drone racing. While domain randomization (DR) improves generalization, single-policy controllers remain constrained by their training distributions and may fail under unmodeled conditions such as wind, hardware wear, or sensor degradation. To address this, we introduce two complementary online switching strategies: the look-back switch, which retrospectively identifies the dynamics model that best matches the current flight state, and the look-ahead switch, which prospectively selects the most suitable policy by simulating candidate controllers over short horizons. Both mechanisms leverage pre-trained policies without requiring retraining, enabling safer and faster flight. We evaluate the methods in simulation on two platforms (5-inch racing quadcopter and the high-fidelity A2RL competition drone) and across two tracks (Figure-8 and A2RL Grand Challenge). Results show that adaptive switching improves robustness and reduces lap times compared to fixed-policy baselines, demonstrating its potential for real-world deployment.},

note = {de Wagter, C. (mentor); Ferede, R. (mentor); Delft University of Technology (degree granting institution); graduation 22-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:57263001-b839-4624-a23b-67cafb8a8441,

title = {Spiking Neural Networks for High-Speed Continuous Quadcopter Control Using Proximal Policy Optimization: Toward Energy-Efficient Neuromorphic Control of Agile Drones},

author = { M.F. van Breukelen},

url = {https://resolver.tudelft.nl/uuid:57263001-b839-4624-a23b-67cafb8a8441},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {We present the first demonstration of a fully spiking actor-critic neural network policy, trained via Proximal Policy Optimization (PPO), for continuous control of an agile high-speed quadcopter in a gate-based navigation task. The spiking neural network (SNN) controller employs Leaky Integrate-and-Fire neurons with surrogate gradient training and spike-rate decoding over multiple integration cycles, and it is benchmarked against a comparable artificial neural network (ANN) controller in both simulation and real-world flight tests. Results show that despite being trained to the same reward level, the SNN achieves superior performance in simulation, achieving higher episode rewards, greater robustness and reduced crash rate. Additionally, in 12-second real-world trials, the SNN outperforms the ANN, attaining a higher average reward (70.63 vs 59.77), greater mean velocity (7.94 vs 6.99 m/s), and more gates cleared (46.33 vs 40.67). An analysis of the spike integration cycle count reveals a clear trade-off: lower cycle counts (fewer integration steps per control update) reduce control output resolution and hinder learning, whereas higher cycle counts improve smoothness but increase inference latency. Moderate cycle counts (5 or 8) provide the best balance, yielding high rewards, smoother outputs, and low execution time overhead. These findings represent a key step forward for neuromorphic control in embedded autonomous systems, demonstrating that SNN-based policies can outperform conventional ANN controllers in high-speed, agile robotic tasks.},

note = {de Wagter, C. (mentor); de Croon, G.C.H.E. (graduation committee member); Ferede, R. (graduation committee member); Vos, R.W. (graduation committee member); Schrama, E.J.O. (graduation committee member); Delft University of Technology (degree granting institution); graduation 19-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:788c2e55-601b-45ae-aa53-8ea45001a3c8,

title = {SPICE: Self-supervised Predictive Coding of Events},

author = { T.N.A. den Blanken},

url = {https://resolver.tudelft.nl/uuid:788c2e55-601b-45ae-aa53-8ea45001a3c8},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Event-based cameras provide high temporal resolution, robustness to lighting conditions and low power consumption, but their sparse, temporal data require models that reason over time. In supervised settings, this is increasingly handled with recurrent architectures. In contrast, most self-supervised learning (SSL) methods still adapt non-recurrent RGB techniques, with masking-based objectives that favor spatial reconstruction over temporal understanding. We introduce SPICE: Self-supervised Predictive Coding on Events, an SSL framework tailored to event data that processes longer sequences recurrently and learns by predicting future latent representations rather than reconstructing masked inputs, promoting a more natural objective focyused on anticipating what comes next. SPICE further incorporates an event-specific contrastive loss only operating on active regions. SPICE pre-training improves downstream performance on semantic segmentation, depth estimation and optical flow estimation. Low-dimensional projections confirm that the learned representations are meaningful and avoid collapse, while also revealing limitations in temporal stability and semantic organization, indicating clear directions for future event-specific SSL research. Code is available upon request.},

note = {Wu, Y. (mentor); de Croon, G.C.H.E. (mentor); Caesar, Holger (graduation committee member); de Wagter, C. (graduation committee member); Ferranti, L. (graduation committee member); Delft University of Technology (degree granting institution); graduation 18-12-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:78972369-0687-445f-9031-a83e10ba600f,

title = {Robust H-infinity Controller Design For INDI-Controlled Quadcopters Using Online Parameter Identification},

author = { T. Aantjes},

url = {https://resolver.tudelft.nl/uuid:78972369-0687-445f-9031-a83e10ba600f},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {This work presents the design of a robust gain-scheduled controller for attitude control of quadcopters, using an Incremental Nonlinear Dynamic Inversion (INDI)-based inner loop with online identification of its system parameters. A linearized model including uncertainty in the identified parameters and unmodeled dynamics is presented for robustness analysis, followed by a set of design requirements. With this, a cascaded feedback attitude controller with a feedforward filter was synthesized for a symmetric quadcopter using signal-based H-infinity closed-loop shaping. Subsequent linear analysis demonstrated good robustness margins and performance characteristics, which were further validated through nonlinear simulations and experimental flights, showing good performance under uncertainty. This methodology was then extended using co-design to develop a gain-schedule for varying actuator time constants. The approach achieved the requirements over the entire range. The resulting gain-scheduled controller exhibited good stability margins, with nonlinear simulations confirming effective tracking performance under uncertainty. Experimental evaluation of the gain-scheduled controller was conducted through flight tests with full online parameter identification. Even though the identified parameters during these tests were far outside the defined uncertainty range, acceptable flight performance comparable to simulation results was maintained for actuator time constants below 40 ms. For slower actuators, performance was degraded but this may have been due to the extreme uncertainties rather than the controller itself.},

note = {Smeur, E.J.J. (mentor); Theodoulis, Spilios (mentor); Blaha, T.M. (mentor); Hamaza, S. (graduation committee member); Bombelli, A. (graduation committee member); Delft University of Technology (degree granting institution); graduation 05-11-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:7addd028-dd34-4d79-8cfa-ccd4ede256c2,

title = {Design, Modeling, and Identification of a Passive Tilting Wing Based on Offset Distribution Electric Propulsion: A Wind Tunnel Based Engineering Approach},

author = { B. Chen},

url = {https://resolver.tudelft.nl/uuid:7addd028-dd34-4d79-8cfa-ccd4ede256c2},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {The increasing demand for sustainable, efficient, and versatile air mobility solutions has accelerated research into Vertical Take-off and Landing (VTOL) configurations. While tilt-rotor and tilt-wing designs offer high cruise efficiency by employing the same propulsion units for lift and thrust, conventional implementations rely on complex and heavy actuation systems, limiting payload capacity and endurance. This thesis proposes a novel passive tilt-wing concept based on offset Distributed Electric Propulsion (DEP), in which asymmetric propeller placement and differential thrust are used to control wing tilt without dedicated actuators. The configuration aims to reduce structural complexity and weight while enhancing aerodynamic performance. However, effective modeling of coupled propeller–wing aerodynamics across a wide AoA range, which is critical for tilt-wing VTOL aircraft, remains lacking. Existing methods either lack accuracy at high AoA (low-order models) or are too computationally expensive for early-stage design (high-fidelity CFD), limiting their applicability to tilt-wing DEP systems. To address this issue, a data-driven aerodynamic modeling framework is developed to predict propeller–wing interactions over a wide angle-of-attack range. The approach combines a physics-informed propeller digital twin—constructed from wind tunnel load and Particle Image Velocimetry (PIV) measurements—with computational fluid dynamics (CFD)-based clean-wing aerodynamic results, corrected for slipstream effects. This model enables accurate and computationally efficient estimation of aerodynamic loads under conditions typical of VTOL transition flight, and is validated against wind tunnel experiments. Parametric studies are conducted to explore the influence of propeller installation parameters on cruise and take-off efficiency, revealing optimal placements for both performance and control authority. The thesis is organized as follows: Part I presents the scientific paper, focusing on the principal innovations of this project in the field of propeller–wing interaction analysis, detailing the development, validation, and application of the modeling of propeller-wing aerodynamic interaction. Part II covers other aspects of the project, including the literature review, the conceptual design of the passive tilt-wing VTOL configuration, preliminary explorations of propeller–wing modeling, and the feasibility analysis and optimization of the proposed passive tilt-wing design.},

note = {Sinnige, T. (mentor); Smeur, E.J.J. (mentor); Delft University of Technology (degree granting institution); graduation 27-08-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:7deb4c84-4804-4bce-a1ec-a2d85afe8ef5,

title = {EAAE: Energy-Aware Autonomous Exploration for UAVs in Unknown 3D Environments},

author = { J.E. Elskamp},

url = {https://resolver.tudelft.nl/uuid:7deb4c84-4804-4bce-a1ec-a2d85afe8ef5},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Autonomous exploration enables unmanned aerial vehicles (UAVs) to map unknown environments without human intervention. While state-of-the-art algorithms primarily focus on maximizing coverage or minimizing exploration time, they often overlook energy efficiency, a critical constraint for battery-powered UAVs. This thesis introduces EAAE, a modular Energy-Aware Autonomous Exploration framework that explicitly incorporates energy consumption into the exploration decision-making process. By combining frontier detection, a K-means divisive clustering algorithm, energy-aware target selection, and dual-layer planning, the framework balances information gain with energy cost. The algorithm is evaluated in simulation using the Agilicious control stack, with a physically realistic power model based on rotor speeds. Two 3D environments of varying complexity are used to compare EAAE against two baselines, an information-gain-based frontier method and distance-based frontier methods. Results show that EAAE consistently achieves the lowest total energy consumption while maintaining competitive or lower exploration times. Moreover, it demonstrates similar performance in terms of information entropy and reduced power variability, contributing to more efficient and robust mapping. This work highlights the importance of integrating energy-awareness into exploration pipelines and provides a foundation for further research into long-endurance autonomous aerial missions.},

note = {Popovic, M. (mentor); de Wagter, C. (graduation committee member); Pek, C. (graduation committee member); Shi, Moji (mentor); Bauersfeld, Leonard (mentor); Delft University of Technology (degree granting institution); graduation 09-07-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:9ab5e5ea-fd2d-434f-9f90-f95c1bed9f1d,

title = {Evolving Behaviour Trees to Control a Swarm of Flapping-Wing Micro Aerial Vehicles for Greenhouse Exploration},

author = { L.H. Uptmoor},

url = {https://resolver.tudelft.nl/uuid:9ab5e5ea-fd2d-434f-9f90-f95c1bed9f1d},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Micro aerial vehicles have shown promising use to further automate food production in greenhouses recently. Compared to conventional multirotor drones, flapping-wing drones offer safe and robust operation around plants due to their soft, slowly-moving wings. Their limited sensing and computational capabilities, however, prohibit the use of map-based navigation methods. To compensate for individual shortcomings, swarming ensures scalability and redundancy. This work proposes a hardware setup combining time-of-flight (ToF) and ultra-wideband (UWB) sensing and explores the artificial evolution of behaviour trees as a reactive planning strategy. Genetic programming, paired with CMAES fine-tuning was able to improve a human-designed exploration strategy by 50%. Neuroevolution has been investigated to encourage emergent swarming behaviours, but requires further experimentation in combination with behaviour trees. The solution obtained in simulation can be readily ported to hardware, but a reality gap in performance persists. These findings contribute to the development of lightweight, scalable aerial systems for autonomous greenhouse monitoring.},

note = {Stroobants, S. (mentor); Popovic, M. (mentor); de Croon, G.C.H.E. (mentor); Delft University of Technology (degree granting institution); graduation 19-09-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:b6872e26-7272-460b-9b13-97e76a9455e0,

title = {Osprey Simulator: A Simulator Framework for Fixed-Wing UAV Updraft Localization},

author = { K.V.A. Vleeschouwer},

url = {https://resolver.tudelft.nl/uuid:b6872e26-7272-460b-9b13-97e76a9455e0},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {he real-world application of Micro Air Vehicles (MAVs) is often constrained by their limited flight range and endurance, primarily due to battery limitations. One way to overcome this challenge is by leveraging naturally occurring vertical air currents, a technique inspired by soaring birds. This paper introduces Osprey Simulator, a framework designed to simulate and test energy-efficient soaring flight strategies for fixed-wing drones. Built on Isaac Sim, Osprey Simulator enables the creation of both randomly generated urban environments and real-world environments, such as the TU Delft campus. These environments are paired with accurate wind field simulations using OpenFOAM, providing a robust platform for studying aerodynamic interactions in diverse settings. This functionality allows for the generation of scalable synthetic datasets, including depth images and wind field information, enabling comprehensive exploration of soaring potential across various structural geometries and wind conditions. Using generated synthetic data, a neural network was trained to predict optimal soaring regions by analyzing depth images and wind field information. The network demonstrates the ability to identify updraft and downdraft regions, enabling more efficient path planning for drones in urban environments. By integrating realistic simulations and advanced predictive models, Osprey Simulator serves as a powerful tool for advancing autonomous soaring and extending the operational range of fixed-wing MAVs.},

note = {de Croon, G.C.H.E. (mentor); Hwang, S. (mentor); Remes, B.D.W. (mentor); Delft University of Technology (degree granting institution); graduation 05-02-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:c70e148b-a3ac-424d-a633-3d0310538064,

title = {Predictability Awareness For Efficient and Robust Multi-Agent Coordination},

author = { R. Chiva Gil},

url = {https://resolver.tudelft.nl/uuid:c70e148b-a3ac-424d-a633-3d0310538064},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {To safely and efficiently solve motion planning problems in multi-agent settings, most approaches attempt to solve a joint optimization that explicitly accounts for the responses triggered in other agents. This often results in solutions with an exponential computational complexity, making these methods intractable for complex scenarios with many agents. While sequential predict-and-plan approaches are more scalable, they tend to perform poorly in highly interactive environments. This paper proposes a method to improve the interactive capabilities of sequential predict-and-plan methods in multi-agent navigation problems by introducing predictability as an optimization objective. We interpret predictability through the use of general prediction models, by allowing agents to predict themselves and estimate how they align with these external predictions. We formally introduce this behavior through the free-energy of the system, which reduces (under appropriate bounds) to the Kullback-Leibler divergence between plan and prediction, and use this as a penalty for unpredictable trajectories. The proposed interpretation of predictability allows agents to more robustly leverage prediction models, and fosters a ‘soft social convention' that accelerates agreement on coordination strategies without the need of explicit high level control or communication. We show how this predictability-aware planning leads to lower-cost trajectories and reduces planning effort in a set of multi-robot problems, including autonomous driving experiments with human driver data, where we show that the benefits of considering predictability apply even when only the ego-agent uses this strategy.},

note = {Alonso-Mora, Javier (mentor); de Croon, G.C.H.E. (mentor); Jarne Ornia, D. (mentor); Khaled Mustafa, K.A. (mentor); de Wagter, C. (graduation committee member); Delft University of Technology (degree granting institution); graduation 14-01-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:cf22cf83-eaaa-4d0c-95ab-0d86aed09987,

title = {Autonomous Aircraft Maintenance Inspection: 3D Frontier-Following Path Planning for UAVs using Delayed Depth Sensing},

author = { L.B. Hofman},

url = {https://resolver.tudelft.nl/uuid:cf22cf83-eaaa-4d0c-95ab-0d86aed09987},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {The use of UAVs for aircraft maintenance inspections has emerged as a promising solution to reduce maintenance costs and increase inspection quality by providing efficient and accurate data collection. In this paper a novel path planning approach is proposed for autonomous aircraft maintenance inspections using UAV swarms, leveraging a frontier-following algorithm. The algorithm guides a drone's trajectory to cover an aircraft's surface while maintaining a constant distance tangent to the surface. Designed to operate with delayed depth sensing methods, where scenes are reconstructed using structure from motion principles, this approach can be integrated with online reconstruction techniques. The frontier-following methodology results in a characteristic spiraling pattern behavior and facilitates complex surface tracking. The primary environment sensing detector, an RGB camera, is angled towards a target on the frontier. This allows for preliminary mapping of a route prior to traversal, while a collision avoidance strategy ensures collision-free flight by tracking an obstacle-free tunnel around the drone. Furthermore, this method can be scaled up for application to swarming agents, accelerating surface coverage. Simulation results demonstrate the algorithm's efficiency in achieving comprehensive surface coverage. Frontier-following path planning outperforms the 2D coverage path planning algorithm Random Walk, while performing lower than Spiral Spanning Tree Coverage - a result consistent with the latter's optimality for 2D surfaces. The proposed method's ability to successfully track complex 3D surfaces, a capability lacking in traditional 2D coverage path planning algorithms, renders it a promising solution for autonomous inspection systems.},

note = {de Wagter, C. (mentor); Henquet, Mathijs (mentor); Borst, C. (graduation committee member); Vanhamel, J.A.M. (graduation committee member); Delft University of Technology (degree granting institution); graduation 11-02-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:d5b09aac-8713-43c6-8bd4-659e4bfe068c,

title = {Neural Inspired Autonomous Drone Racing: Towards a modular, competitive framework},

author = { E. Lucassen},

url = {https://resolver.tudelft.nl/uuid:d5b09aac-8713-43c6-8bd4-659e4bfe068c},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {The goal of this thesis was to push the boundaries of fast and agile flight for autonomous drone racing. A framework was developed for high-performance vision-based localization, sensor fusion, and control, with a focus on integration of neural networks. Additionally, a neural dynamics model was created which can can replace all accelerometer measurements by thrust&drag predictions, while running directly on the flight controller.},

note = {de Wagter, C. (mentor); de Croon, G.C.H.E. (graduation committee member); Delft University of Technology (degree granting institution); graduation 21-03-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:d8c0486d-f7eb-443a-8c76-9770038db3e8,

title = {Reinforcement learning driven behavioral entropy based frontier exploration},

author = { C. Li},

url = {https://resolver.tudelft.nl/uuid:d8c0486d-f7eb-443a-8c76-9770038db3e8},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Autonomous robotic exploration must balance fast mapgrowth with robustness under uncertainty. Among frontier-based methods, information-theoretic variants (e.g., Shannon/Rényi) are fast and reliable. This thesis integrates Behavioral Entropy (BE)—which models human-like risk perception—into a planner-in-theloop deep reinforcement learning(DQN) framework that learns the BE risk parameter 𝛼 online. The policy selects 𝛼 from a discrete action set, uses occlusion-aware visibility to compute BE-based information gain, and delegates motion to a classical A* planner; reward shaping couples coverage/entropy reduction with motion cost. Action-usage analysis reveals an adaptive risk schedule—aggressive early, conservative mid-episode, selectively aggressive late—enabling faster cleanup of residual area. Overall, a reinforcement-learning method with risk-attitude tuning yields a robust, planner-compatible explorer.},

note = {Popovic, M. (mentor); van Kampen, E. (mentor); de Pater, I.I. (mentor); Delft University of Technology (degree granting institution); graduation 30-10-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:f4f3dc1b-26f2-4107-8c28-8baf1451be6c,

title = {Autonomous Navigation for an Attitude-Stable Flapping Wing Air Vehicle: Obstacle Avoidance Strategies Using Time-of-Flight Sensors},

author = { S. Blaga},

url = {https://resolver.tudelft.nl/uuid:f4f3dc1b-26f2-4107-8c28-8baf1451be6c},

year  = {2025},

date = {2025-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {In recent years, flapping wing micro aerial vehicles (FWMAVs) have garnered significant attention due to their agility in cluttered environments and the safety advantages offered by their soft wings during close-proximity operations. These vehicles are subject to numerous constraints: limited power supply, restricted payload capacity, high drag and vibration from the flapping mechanism, and susceptibility to environmental disturbances such as wind gusts—making indoor operation preferable. However, these vehicles have not achieved autonomous navigation yet. The platform used in this study, the Flapper Nimble+, is attitude-stable, meaning it lacks access to egomotion or positional information and is only capable of controlling its thrust, pitch, roll, and yaw. Given the strict SWaP (Size, Power and Weight) constraints, lightweight Time-of-Flight (ToF) sensors are among the most practical sensing solutions. This thesis investigates the question: How can obstacle avoidance be achieved in an attitude-stable flapping wing air vehicle equipped with Time-of-Flight sensors? Two control approaches were implemented and tested in the IsaacGym simulator: a simple PID controller with confidence-based yaw adjustment, and a reinforcement learning (RL) policy trained using Proximal Policy Optimization. Two sensor configurations were considered: a minimal two-sensor setup (front and downward) and a richer five-sensor array for broader perception. The PID controller successfully avoided collisions in all environments, demonstrating high reliability but limited exploration, averaging 35% coverage. In contrast, the RL policy with five sensors achieved greater spatial exploration—averaging 49.5% coverage—at the cost of an average of 3.5 collisions per episode. When reduced to two sensors, the RL agent’s performance declined significantly, with average coverage dropping to 29% and collision counts increasing to 6.6 (excluding failed runs). These results show that autonomous navigation is achievable for attitude-stable flapping wing air vehicles using only ToF sensors. While PID control offers reliable, conservative navigation under minimal sensing, RL enables more exploratory behaviors and adaptability in dynamic environments.},

note = {de Croon, G.C.H.E. (mentor); de Wagter, C. (mentor); Delft University of Technology (degree granting institution); graduation 14-05-2025},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:c30dd07b-d2ca-4fed-98c9-7f4b874c7f26,

title = {STEM: Semantic Target Search and Exploration using MAVs in Cluttered Environments},

author = { N. Sethi},

url = {https://resolver.tudelft.nl/uuid:c30dd07b-d2ca-4fed-98c9-7f4b874c7f26},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Mechanical Engineering},

abstract = {Autonomous target search is crucial for deploying Micro Aerial Vehicles (MAVs) in emergency response and rescue missions. Existing approaches either focus on 2D semantic navigation in structured environments – which is less effective in complex 3D settings, or on robotic exploration in cluttered spaces – which often lacks the semantic reasoning needed for efficient target search. This thesis overcomes these limitations by proposing a novel framework that utilizes semantic reasoning to minimize target search and exploration time in unstructured environments using a MAV. Specifically, the open vocabulary inference capabilities of Large Language Models are employed to embed semantic relationships in segmentation images. An active perception pipeline is then developed to guide exploration toward semantically relevant regions of 3D space by biasing frontiers and selecting informative viewpoints. Finally, a combinatorial optimization problem is solved using these viewpoints to create a plan that balances information gain with time costs, facilitating rapid location of the target. Evaluations in complex simulation environments show that the proposed method consistently outperforms baselines by quickly finding the target while maintaining reasonable exploration times. Real-world experiments with a MAV further demonstrate the method's ability to handle practical constraints like limited battery life, small sensor range, and semantic uncertainty.},

note = {Lodel, M. (mentor); Alonso-Mora, Javier (mentor); Popovic, M. (graduation committee member); Hernández, Carlos (graduation committee member); Delft University of Technology (degree granting institution); graduation 27-08-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:10c5b8e1-cc6c-4537-b341-7c084c16ef4f,

title = {Autonomous Aerial Sensor Network Placement in Rainforests: Exploration and Detection of Deployment Locations},

author = { R. Santos Raminhos},

url = {https://resolver.tudelft.nl/uuid:10c5b8e1-cc6c-4537-b341-7c084c16ef4f},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {The effects of climate change put increasing strain on rainforests and their inhabitants, highlighting the demand for technological developments to aid in biodiversity monitoring and conservation efforts. This research work proposes a novel frame work for the autonomous placement of acoustic sensor networks on top of the rainforest canopy, using a quadrotor platform. In order to tackle the challenge posed by multirotors’ limited autonomy, an initial canopy exploration mission is considered and an exploration planner is developed to detect suitable locations on top of the rainforest canopy for sensor node deployment. Specifically, green detection and pointcloud projection is performed online and combined with our proposed sampling method to accomplish targeted exploration, towards the estimated location of detected green components. Flight experiments are conducted across multiple scenarios that mimic distinct rainforest features. The results validate the system architecture and demonstrate the effectiveness of our smart sampling method, laying the foundation for future autonomous sensor network exploration missions at larger-scale.},

note = {Hamaza, S. (mentor); de Croon, G.C.H.E. (graduation committee member); Rajan, R.T. (graduation committee member); Delft University of Technology (degree granting institution); graduation 26-11-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:3cbf5a3f-4929-464c-ae7a-e28b876dc964,

title = {Reinforcement Learning for Spiking Neural Networks: Recurrent Reinforcement Learning with Surrogate Gradients},

author = { K.C.M. van den Berghe},

url = {https://resolver.tudelft.nl/uuid:3cbf5a3f-4929-464c-ae7a-e28b876dc964},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Enabling embodied intelligence in robotics presents several unique challenges. A first major concern is the need for energy efficiency, low latency, and strong temporal reasoning to facilitate effective real-world interaction. Neuromorphic computing has garnered attention as a potential solution to these problems. Secondly, when using deep neural networks, it is hard to shape a learning signal, due to the goal oriented nature of robotics. Reinforcement learning (RL) poses itself as a framework to leverage goal-directed reward functions to create this learning signal. A key challenge with recurrent and spiking neural networks trained via RL is achieving stable baseline performance, able to creating sequences long enough to stabilize hidden states. This stabilization is crucial for processing sequences that extend beyond the initial warm-up period of the temporal network. In this article, an online RL approach is proposed, enabling temporal training with minimal changes to existing online algorithms, introducing a secondary guiding policy whose sole objective is to prevent episode termination before the warm-up period is complete. This framework is demonstrated to outperform offline RL methods and significantly improve the wall clock time of online RL methods, adapted to sample sequences rather than single transitions. Next, the effect of surrogate gradients as a technique for translating the learning signal from the RL framework to weight updates is analyzed. It is found that the slope, parametrizing the surrogate gradient, plays a crucial role in online RL settings, and can be exploited as an exploration mechanism.},

note = {de Croon, G.C.H.E. (mentor); Stroobants, S. (mentor); de Wagter, C. (graduation committee member); Zarouchas, D. (graduation committee member); Delft University of Technology (degree granting institution); graduation 19-12-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:411c4c11-8401-48e1-8977-4051d81c27f9,

title = {Drones' Tactile Navigation using Biomimetic Vibrissal Sensors},

author = { M.H. Yoganarasimhan},

url = {https://resolver.tudelft.nl/uuid:411c4c11-8401-48e1-8977-4051d81c27f9},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {This study explores the potential of biomimetic vibrissal sensors for tactile navigation in aerial robotics. Inspired by the sophisticated sensory system of rodents, we developed a non-intrusive, whisker-based tactile sensor system integrated into a drone platform. The system enables real-time detection and navigation through complex environments by emulating the tactile perception capabilities of natural whiskers. Our approach includes a novel platform design for easy sensor integration, a preprocessing solution to mitigate signal distortion as well as a simple static calibration set-up to estimate normal contact distances. The effectiveness of the system was validated through contour following tasks, where the whisker sensors provided feedback for precise navigation along surfaces with varying orientations. Results demonstrate that our system can estimate normal distances and wall orientations with sufficient accuracy, despite challenges such as lateral slip. This research highlights the potential of whisker-inspired sensors in enhancing the tactile sensing capabilities of aerial robots, offering significant advantages over traditional tactile sensors in terms of weight, power consumption, and operational flexibility in diverse environmental conditions.},

note = {Hamaza, S. (mentor); Masania, K. (graduation committee member); Delft University of Technology (degree granting institution); graduation 11-09-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:4fcc12d8-28dd-4d92-9a8f-a32b5554fcd6,

title = {Hunt like a Dragonfly and Strike like a Drone: Optimizing quadcopter control for insect pest interception through multi-agent deep reinforcement learning},

author = { R.W. Vos},

url = {https://resolver.tudelft.nl/uuid:4fcc12d8-28dd-4d92-9a8f-a32b5554fcd6},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Insect pest elimination through MAV interception can reduce the need for insecticides and can contribute to sustainable agriculture. In this research, we analyze the feasibility of such solutions through simulated two-player differential games of pursuit and evasion with agents operating on minimalistic sets of biologically-plausible observations and optimized to control constrained vehicle models through deep multi-agent reinforcement learning. Our pursuer and evader agent, representing the quadcopter drone and insect pest respectively, are asymmetric in design, capabilities and objectives. Our results show that our quadcopter pursuer is consistently able to pursue and intercept a reactive insect-inspired evader as well as recordings of actual insect targets, achieving interception rates of 55% and 94% on these respective tasks. In comparison, pursuers alternatively optimized against non-reactive evaders or reactive drone-like evaders with symmetric capabilities, achieve an interception rate of only 42% for the same insect target recordings. Despite these promising results, we conclude that further research is needed to formally establish the superiority of multi-agent optimization in this asymmetric game scenario. Finally, we determine how emergent behavior and strategies resemble nature. During the confrontations, we observe that our pursuer mainly implements pure-pursuit as well as motion camouflage to some degree; drawing comparison to the hunting strategy of dragonfly.},

note = {de Croon, G.C.H.E. (mentor); Yedutenko, M. (mentor); Boyle, J.H. (graduation committee member); de Wagter, C. (graduation committee member); Delft University of Technology (degree granting institution); graduation 04-10-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:578aa4a3-83cc-4f89-92d2-b68e4fd081d7,

title = {Insect inspired visual homing via direct home vector prediction - a simulation study},

author = { M.R. Jankowski},

url = {https://resolver.tudelft.nl/uuid:578aa4a3-83cc-4f89-92d2-b68e4fd081d7},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {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.},

note = {de Croon, G.C.H.E. (mentor); Hagenaars, J.J. (mentor); Smeur, E.J.J. (graduation committee member); Verhoeven, C.J.M. (graduation committee member); Delft University of Technology (degree granting institution); graduation 05-11-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:6d26bb58-9188-42eb-bccc-b8a823c538d9,

title = {Autonomous Ceiling-Following Using Biomimetic Vibrissal Sensors on a Quadrotor},

author = { N.W.M. de Krom},

url = {https://resolver.tudelft.nl/uuid:6d26bb58-9188-42eb-bccc-b8a823c538d9},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Low visibility in hard-to-reach environments, such as natural caves and tunnels, poses challenges for exploration due to the limited effectiveness of traditional sensors, such as cameras and distance sensors. Nature offers a solution to this problem embedded in the body of nocturnal animals in the form of vibrissal sensors, whiskers. In this study, we take inspiration from this principle and develop artificial biomimetic whiskers to aid quadrotors' navigation. Specifically, we design and control a whiskered drone for ceiling contour-following tasks, enabling tactile tracing of rugged ceilings while mitigating aerodynamic disturbances like the "ceiling effect". The proposed whiskered drone design allows for multiple whiskers in various orientations, providing real-time tactile feedback to guide the ceiling-following controller. Flight experiments validate the system's effectiveness in navigating both smooth and rugged ceilings, demonstrating the potential for whisker-based tactile navigation and future autonomy.},

note = {Hamaza, S. (mentor); Jovanova, J. (graduation committee member); Delft University of Technology (degree granting institution); graduation 24-10-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:7e2a3aff-c782-45c2-b494-929cd40d7c8f,

title = {Reconfigurable Force-Balanced Aerial Manipulator for Variable Payload Tasks},

author = { M. Bianconi},

url = {https://resolver.tudelft.nl/uuid:7e2a3aff-c782-45c2-b494-929cd40d7c8f},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {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.},

note = {Hamaza, S. (mentor); Smeur, E.J.J. (graduation committee member); Varriale, Carmine (graduation committee member); Brummelhuis, M.B.J. (mentor); Delft University of Technology (degree granting institution); graduation 04-12-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:83578c70-f202-42f8-8123-092f37621d11,

title = {Insect-Inspired Navigation: A Real-World Drone that Homes Like Honeybees After Foraging Flight},

author = { D. Ou},

url = {https://resolver.tudelft.nl/uuid:83578c70-f202-42f8-8123-092f37621d11},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Insects like honeybees exhibit remarkable navigational abilities despite their simple nervous systems, showcasing expertise in tasks such as long-distance travel, landmark recognition, and spatial memory. These skills are crucial for efficient foraging and homing. In robotics, one of the main challenges is to navigate in GPS-denied environments with limited sensors and processors onboard. In this study, we propose a novel navigation strategy that uses learning flights during which a robot directly maps images to nest location vectors, inspired by honeybees. In our previous research, the learning phase was modeled using compact convolutional neural networks (CNNs) and demonstrated successful learning and control in simulation. In this work, we combine this homing model with odometry and implement both in a real-world quadrotor. More specifically, we examine how this model can compensate for the drone’s odometric drift to reach home after it performs a long-distance outbound flight. Real-world experiments demonstrate the proof of concept of the proposed navigation strategy in both indoor and outdoor flights.},

note = {Hagenaars, J.J. (mentor); de Croon, G.C.H.E. (mentor); Delft University of Technology (degree granting institution); graduation 30-08-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:920aa240-24a5-4625-9a43-6eeac7f82bee,

title = {Learning Generalizable Policy for Obstacle-Aware Autonomous Drone Racing},

author = { Y. Liu},

url = {https://resolver.tudelft.nl/uuid:920aa240-24a5-4625-9a43-6eeac7f82bee},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {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.},

note = {Yu, H.Y. (mentor); de Wagter, C. (mentor); Smeur, E.J.J. (graduation committee member); Hulshoff, S.J. (graduation committee member); Delft University of Technology (degree granting institution); graduation 25-11-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:9b7ad90f-66be-436d-aa96-007277aef528,

title = {Guidance and Control Implementation with Spiking Neural Networks: A feasibility study},

author = { T. Avarvarei},

url = {https://resolver.tudelft.nl/uuid:9b7ad90f-66be-436d-aa96-007277aef528},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {Quadrotors have continuously leveraged the use of artificial intelligence for navigation and decision-making. Moreover, neuromorphic computing, specifically Spiking Neural Networks (SNNs), is considered as an energy-efficient solution during inference. The current study will analyse the effects of implementing SNNs for mimicking energy optimal guidance and control. To achieve this, population encoding is used and an equivalent of 7-8 spiking neurons per conventional neuron is found to preserve most of the information. The equivalent controller prefers fast adaptation which requires small spiking threshold values and minimal reliance on past information. To improve the controller performance, dataset selection is of utmost importance with a careful trade-off between excessive race track customisation and generalisability being required. The results show that learning is feasible and SNN performance approaches conventional state-of-the-art models trained with multi-layer perceptrons. The current analysis represent an important step towards the rapid guidance and control of ultra-small energy efficient quadrotors.},

note = {de Wagter, C. (mentor); Stroobants, S. (mentor); Ferede, R. (graduation committee member); Delft University of Technology (degree granting institution); graduation 18-07-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}

@mastersthesis{uuid:ea79f535-988a-4b92-9740-c6d183cb1f52,

title = {Reinforcement Learning for Landing the Variable Skew Quad Plane on a Moving Platform: Achieving Optimal Guidance for Ship Landings},

author = { C.Y. YIKILMAZ},

url = {https://resolver.tudelft.nl/uuid:ea79f535-988a-4b92-9740-c6d183cb1f52},

year  = {2024},

date = {2024-01-01},

school = {TU Delft Aerospace Engineering},

abstract = {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.},

note = {de Wagter, C. (mentor); Smeur, E.J.J. (graduation committee member); De Domenico, F. (graduation committee member); Delft University of Technology (degree granting institution); graduation 14-08-2024},

keywords = {},

pubstate = {published},

tppubtype = {mastersthesis}

}