Recent advances in AI, driven by increased computational power, are expanding capabilities in aerospace. While machine learning (ML) reveals hidden data patterns, it typically requires large datasets. Numerical simulations offer insights but are costly and sometimes unreliable due to model limitations. Experimental testing, though essential, is expensive and yields limited data. Physics-Informed Neural Networks (PINNs) integrate physical laws into ML models, reducing data needs and improving accuracy. This study presents a PINN tool developed in Python with PyTorch, supporting flexible architectures and adaptable cost functions. PINNs were applied to aerospace use cases, including temperature prediction, computational fluid dynamics (CFD), and Solid Rocket Motor modelling. Compared to traditional numerical and purely data-driven methods, PINNs demonstrated improved accuracy and data efficiency, though with higher computational costs. They also show promise as surrogate models. Future work will focus on optimising for dedicated hardware, deeper architectures, and broader applications to further explore their potential in aerospace simulation and modelling.

The NIBIRU mission, scheduled for launch in 2038, seeks to image and characterise Planet 9, a hypothesized planet beyond Neptune in the Kuiper Belt. The mission will travel farther than any human-made object and involves a close approach to the Sun, necessitating precise design and integration of all subsystems. This report details the mission's objectives, design processes, subsystem specifics, and risk assessments. Primary objectives include confirming Planet 9's existence and location, estimating its mass and radius, capturing detailed images, identifying surface and atmospheric features, detecting moons or rings, determining atmospheric composition, analysing oxygen isotope ratios, and identifying potential biosignatures. Secondary objectives involve characterizing planets used for gravity assists, studying Kuiper Belt objects, and analysing the boundaries of the heliosphere. A comprehensive trade-off analysis evaluated four mission concepts based on criteria like scientific objectives, risk, cost, communications, flexibility, and sustainability. Concept C1, featuring a spacecraft that will perform in-situ measurements and communicate directly with Earth, was selected for its high performance and feasibility. Subsystem designs were meticulously developed. The payload includes instruments such as the N'LORRI imager, ISHTAR imaging spectrometer, NCREX cosmic ray telescope, and PSP particle science package. The Electrical Power Subsystem (EPS) relies on two eMMRTGs and additional batteries for power. The Telecommunications subsystem features a large deployable antenna for long-distance communication. The Attitude Determination and Control System (ADCS) uses spin stabilization and thrusters for precise control, as well as star trackers and IMUs. The propulsion system is a liquid bipropellant system using nitrogen tetroxide and monomethylhydrazine. The Thermal Control System (TCS) combines multi-layer insulation and ceramic carbon tiles to manage extreme temperatures. Structural components use aluminium alloys for strength and durability. System integration ensured compliance with mission requirements, with a total spacecraft mass of 40,648 kg and mission cost of €M3441. Risk assessments identified and mitigated 114 risks, with detailed contingency plans in place. The mission's sustainable development strategy emphasizes the need to use reusable launchers, green energy, and sustainable off-the-shelf materials. The detailed design, manufacturing, testing, and integration phases are meticulously planned to ensure mission readiness for a 2038 launch, promising significant contributions to our understanding of the outer solar system.

Airlab, a collaboration between TU Delft and Ahold Delhaize, is developing Albert, a robot tailored to work in a complex supermarket environment. Key to Albert is a product detection and classification module that tells it what products to grasp and where they are located in a shelf. Albert’s existing YOLO‑based product detector a significant issue: Adding new products without re‑training the whole model is impossible. Especially in a dynamic supermarket environment with an ever‑changing stock, the latter is a major issue.

This problem will be the main focus of this paper and is addressed through few‑shot learning, which predicts similarity between query and target products. This simplifies adding new products to just supplying new target images. Few‑shot learning also requires significantly less data to train on. In supermarkets with 300.000 different products, requiring only a few images per product is a major advantage. For this reason, this paper aims to deploy a few‑shot model to classify products as either the target class or non‑target class for Albert’s picking task and defines the following research question: “What few‑shot classifier can identify products in a supermarket environment, is able to detect non‑target classes, and meets the requirements of deployment on a robotic platform like Albert best?“

This paper first analyses the potential of using TRIDENT and P>M>F, two state‑of‑the‑art few‑shot models, for deployment on Albert, and evaluates them on the requirements of this paper. P>M>F performs better on all requirements, which makes it the preferred model for Albert. However to work well, it still requires adjustments. Its inference time is still too high to work on Albert and it cannot classify query images as not the target product.

For this reason, this paper uses P>M>F’s two key ideas to construct Product‑ProtoNet, a new Albert‑suitable few‑shot model: 1) Using a good pre‑trained feature extractor; and 2) Comparing query images to a set of classes and matching only to the likeliest. P>M>F uses a ProtoNet model for classification that essentially does this; Like ProtoNet, Product‑ProtoNet constructs class prototypes from one or multiple examples of class images. Product‑ProtoNet then uses a sigmoid classifier to predict if query images have the same class as those prototypes. It compares query images to a set of similar class prototypes(helper prototypes) and classifies it as the likeliest. Product‑ProtoNet uses a ViT pre‑trained with DINO to extract image features. To bring down inference time, Product‑ProtoNet computes product prototypes before deployment.

With an accuracy of 99.1% on product classes seen during training and 99.8% on novel classes in a realistic supermarket setting, a low inference time of 2.89 ms and a memory usage lower than 4GB, Product‑ProtoNet is the only model that passes all requirements of this paper. When deployed on Albert Product‑ProtoNet successfully guides Albert to the right product in 97% of attempts. This makes Product‑ProtoNet the only few‑shot classifier that can identify products in a supermarket environment, is able to detect non‑target classes, and meets the requirements of deployment on a robotic platform like Albert.