Ikerlan

Large Language Models are increasingly used for high-level robotic reasoning, yet their latency and stochasticity complicate their direct use in low-level control. Moreover, extracting actionable navigation cues from multimodal context incurs inference costs that are challenging for embedded platforms. We present a plug-and-play framework that augments a finite-state machine with asynchronous velocity suggestions generated by a Large Language Model, using an off-the-shelf DistilGPT-2 model running on-device on a Jetson AGX Orin. The system extracts task-relevant cues from the current context and integrates them only if they satisfy deadline, schema, and kinematic validation, thereby preserving a deterministic 50 Hz control loop with a <5 ms fallback path. We compare multiple Large Language Models for embedded robot control and quantify trade-offs among model size, inference time, and output validity. To assess whether the Large Language Models add value beyond signal processing, we include an ablation against a standard smoothing baseline; the results indicate that the Large Language Models contribute anticipatory, context-dependent adjustments that are not captured by filtering alone. Experiments in Gazebo and on a real TurtleBot3 reduce the final position error from 0.246 m to 0.159 m and improve trajectory efficiency from 0.821 to 0.901 without increasing control-loop latency. Approximately 80% of the Large Language Models’ outputs pass validation and are applied. Overall, the framework reduces developer effort by enabling behavioral changes at the prompt level while maintaining interpretable, robust edge-based navigation.

Next-generation transportation and industrial systems are characterized by a continuously increasing number of integrated software functionalities and computational performance requirements. Nevertheless, there is also a need to reduce the cost, number, size, power consumption, volume and weight of integrated dependable embedded systems. In this context, virtualization technology enables the reconciliation of these challenges supporting the integration of multiple software functionalities of mixed-criticality (real-time and non-real-time, safety and non-safety-related) on a single computing device. This survey provides an overview of virtualization solutions (hypervisors, microkernels, and separation kernels) and categorizes research contributions addressing virtualization, real-time, dependability, and domain-specific technological challenges.

Abstract Negative particles can target non-clausal constituents, giving rise to a number of configurations in which the negation particle immediately precedes the negated term. One well-known instance are sequences in which negation precedes a quantifier. Different ways of analysing these sequences have been proposed. A relatively straightforward way is to interpret these sequences as involving Merge of the negation particle and the Quantifier Phrase. But one of the first attempts to make sense of the distribution of these sequences (Lasnik 1972), claimed that the sequences follow a rebracketing operation that involves a clause external negation and a fronted quantifier phrase. We propose, on the basis of data coming from Spanish and Basque, that those two alternatives are needed, and provide an up-to-date version of Lasnik’s early analysis. Languages may use different means to delimit the relative scope of negation and QPs, and Direct Merge between negation and the quantifier phrase is just unattested for many languages. We claim that Negation and QPs can combine by two means: (i) Association with Focus, generally available in all languages, that may or may not involve fronting to a position adjacent to a high, clause-external negation; (ii) Direct Merge of Negation and QP. The two options are shown to be productively available in Spanish. Basque, on the other hand, resorts to association with focus as the general means to construct subclausal negation, even when the relevant constructions show adjacency between negation and the quantifier phrase.

En este artículo se analiza pormenorizadamente el impacto en la mejora de la resiliencia de un aerogenerador que ya opera con un algoritmo tolerante a fallos, mediante la modificación de dos de los parámetros habituales en el controlador como son el par máximo y el ángulo de pitch mínimo. Los resultados obtenidos a partir de las 5.472 simulaciones realizadas con el aerogenerador de referencia DTU 10 MW, demuestran que es factible evitar todas las paradas provocadas por varias fuentes de fallo habituales en el sistema de pitch hidráulico, como las fugas y la pérdida de presión de pre-carga en los acumuladores. Esto conlleva reducciones importantes en la mayoría de las cargas de fatiga mecánica consideradas, con mejoras máximas que superan el 20% en comparación con los resultados obtenidos con los valores nominales. La reducción de las paradas también redunda en un aumento de la producción eléctrica, llegando hasta el 63,8% de mejora en el caso del exceso de fricción. En contraprestación, y en ausencia de fallos, la modificación de los parámetros acarrea una reducción en la potencia generada. Esta pérdida puede verse limitada mediante un conjunto adecuado de valores. La combinación presentada en este artículo minimiza esa pérdida a solo el 5,2%.

Dynamic Vision Sensors (DVS) provide low-latency, high-dynamic-range motion estimation, but their real-time applicability is often limited by the computational complexity and latency overheads introduced by iterative motion compensation techniques. In this work, we propose a novel probabilistic model that leverages the stochastic distribution of events along moving edges. Using our model, we introduce a lightweight patch-based algorithm that employs a linear combination of event spatial coordinates, making it highly suitable for implementation on specialized hardware. Our approach exhibits linear scalability with dimensionality, making it suitable for emerging event-based 3D sensors, such as Light-Field DVS (LF-DVS). Experimental results validate the efficiency and scalability of our method, establishing a solid foundation for real-time event-based ultra-efficient 2D and 3D motion estimation.

This paper presents the results of the H2SAREA project, which focuses on integrating hydrogen (H2) into the existing natural gas (NG) distribution network, with blends of up to 20%. A key component of the project was the H2Loop testing platform, built using ex-service materials and components to realistically assess the impact of hydrogen on current systems and components. The investigation covered several critical areas, including gas injection and blending, network capacity, leak detection, gas pressure regulation station (GPRS) performance, valve and meter functionality, materials compatibility, permeation testing, and gas deblending. Results show the feasibility of safely injecting up to 20% hydrogen into the existing system, offering valuable insights to guide the transition of gas distribution networks toward a hydrogen-based energy future. • Large scale distribution facility (H2Loop) operated successfully for 12,000+ hours with up to 20% H2 in natural gas. • Network confirmed capable of handling 20% hydrogen-natural gas mixtures. • Suitability of deblending technology, materials, components and domestic appliances. • Project supports repurposing gas infrastructure for up to 20% hydrogen. • Findings advocate hydrogen as a key component in future energy strategies.

Maritime transport plays a major role in greenhouse gas emissions, which has prompted initiatives such as the International Maritime Organisation's target to reduce CO2 emissions by 40% by 2030. Electrification of maritime transport presents a promising solution, evidenced by the increasing adoption of hybrid and electric vessels. Managing the charging process and port-specific variables, however, presents challenges. This paper proposes a cloud-based Digital Twin model for optimizing multi-megawatt charging control strategies for electric vessels. The charger model integrates transformer and converter models, crucial for simulating electrical and thermal behaviours. An electrothermal equivalent circuit model accurately represents battery dynamics. The developed control strategy, divided into upper and lower layers, optimizes energy management and vessel charging processes. The cloud-based implementation utilizes Amazon Web Services, integrating the DT model into a scalable and flexible environment. Simulation tests validate the accuracy and effectiveness of the DT model and cloud architecture. Finally, the paper discusses the road to standardization, highlighting the importance of ship-port communications and providing recommendations for future energy control solutions at port authorities.