Abstract
Autonomous landing for Unmanned Aerial Vehicles (UAVs) requires both precision and resilience against environmental uncertainties, capabilities that current approaches struggle to deliver. This paper presents a novel learning-based solution that combines an advanced multimodal transformer-based detector with a reinforcement learning formulation to achieve reliable autonomous landing behavior across varying scenario uncertainties. Beyond the integration of multimodality for robust target detection, this research incorporates a comprehensive analysis of the impact of state representation on decision-making performance. The proposed methodology is validated through extensive simulation studies and real-world field experiments conducted on physical UAV platforms under natural wind disturbances, demonstrating reliable transfer from simulated training environments to controlled outdoor conditions. Field experiments across varying initial conditions and wind stress confirm the system’s robustness, achieving landing precision of 0.10 ± 0.08 meters in outdoor trials, demonstrating centimeter-level accuracy that surpasses the meter-level precision of global positioning systems.

Abstract
State-of-the-art symbolic music generation models have recently achieved remarkable output quality, yet explicit control over compositional features, such as tonal tension, remains challenging. We propose a novel approach that integrates a computational tonal tension model, based on tonal interval vector analysis, into a Transformer framework. Our method employs a two-level beam search strategy during inference. At the token level, generated candidates are re-ranked using model probability and diversity metrics to maintain overall quality. At the bar level, a tension-based re-ranking is applied to ensure that the generated music aligns with a desired tension curve. Objective evaluations indicate that our approach effectively modulates tonal tension, and subjective listening tests confirm that the system produces outputs that align with the target tension. These results demonstrate that explicit tension conditioning through a dual-level beam search provides a powerful and intuitive tool to guide AI-generated music. Furthermore, our experiments demonstrate that our method can generate multiple distinct musical interpretations under the same tension condition.

Abstract
Autonomous Vehicles aim to understand their surrounding environment by detecting relevant objects in the scene, which can be performed using a combination of sensors. The accurate prediction of pedestrians is a particularly challenging task, since the existing algorithms have more difficulty detecting small objects. This work studies and addresses this often overlooked problem by proposing Multimodal PointPillars (M-PP), a fast and effective novel fusion architecture for 3D object detection. Inspired by both MVX-Net and PointPillars, image features from a 2D CNN-based feature map are fused with the 3D point cloud in an early fusion architecture. By changing the heavy 3D convolutions of MVX-Net to a set of convolutional layers in 2D space, along with combining LiDAR and image information at an early stage, M-PP considerably improves inference time over the baseline, running at 28.49 Hz. It achieves inference speeds suitable for real-world applications while keeping the high performance of multimodal approaches. Extensive experiments show that our proposed architecture outperforms both MVX-Net and PointPillars for the pedestrian class in the KITTI 3D object detection dataset, with 62.78% in $AP_{BEV}$ (moderate difficulty), while also outperforming MVX-Net in the nuScenes dataset. Moreover, experiments were conducted to measure the detection performance based on object distance. The performance of M-PP surpassed other methods in pedestrian detection at any distance, particularly for faraway objects (more than 30 meters). Qualitative analysis shows that M-PP visibly outperformed MVX-Net for pedestrians and cyclists, while simultaneously making accurate predictions of cars.

Abstract
Climate change has led to the need to transition to clean technologies, which depend on an number of critical metals. These metals, such as nickel, lithium, and manganese, are essential for developing batteries. However, the scarcity of these elements and the risks of disruptions to their supply chain have increased interest in exploiting resources on the deep seabed, particularly polymetallic nodules. As the identification of these nodules must be efficient to minimize disturbance to the marine ecosystem, deep learning techniques have emerged as a potential solution. Traditional deep learning methods are based on the use of convolutional layers to extract features, while recent architectures, such as transformer-based architectures, use self-attention mechanisms to obtain global context. This paper evaluates the performance of representative models from both categories across three tasks: detection, object segmentation, and semantic segmentation. The initial results suggest that transformer-based methods perform better in most evaluation metrics, but at the cost of higher computational resources. Furthermore, recent versions of You Only Look Once (YOLO) have obtained competitive results in terms of mean average precision.

Abstract

Abstract
In recent years, non-residential buildings have increasingly adopted renewable energy generation systems to align with the European Union's goal of achieving carbon neutrality by 2050. However, energy storage systems play a fundamental role in maximising the use of the generated renewable energy. Due to their high acquisition costs, adequately sizing these systems is essential. Moreover, applying an optimal scheduling strategy for energy storage operation can significantly improve the economic viability of such systems by reducing energy-related costs. In this paper, a MILP-based optimisation algorithm-incorporating battery lifespan constraints-is applied to a reference commercial building to schedule the operation of the storage system. A sensitivity analysis on the installed photovoltaic power and energy storage capacity is performed to evaluate their impact on the economic and operational performance of the optimisation algorithm under different sizing configurations.