87 results
Filters
Settings
Now showing1 - 10 of 87
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , When the swarm changes its mind: multi-task learning-based anomaly detection in decentralized AUV missionsPublicationtype: Conference Paper4 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Demo: sd-ahoi — an open source software-defined modem for acoustic underwater communicationSoftware-Defined Acoustic Modems (SDAMs) are powerful tools for developing, prototyping, evaluating, and comparing novel physical, link, layer and network layer protocols tailored to the specific challenges of acoustic underwater communication. We introduce sd-ahoi, an accessible open source SDAM compatible with GNU Radio and based on the hardware of the open source ahoi modem. We present our design goals and provide a detailed description of the architecture of sd-ahoi. In our demonstration, we transmit text messages between two modems using sd-ahoi while interactively switching between different physical and link layer implementations at runtime.Publicationtype: Conference Presentation34 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Benchmarking acoustic underwater sensor-node localization against seafloor photogrammetry(IEEE, 2026-09); ; ; Gonzáles, EveradoAcoustic ranging is the established approach for localizing seafloor sensor nodes and generally achieves high precision. However, performance evaluation typically relies on the same acoustic ranges and GNSS positions used during localization, making error estimates self-referential. We address this circularity by introducing a photogrammetric benchmark, georeferenced by a ship-based USBL system, that is constructed independently of the acoustic ranges and GNSS positions under evaluation. For a five-node acoustic sensor network deployed at 60m depth offshore Santorini, Greece, we build a georeferenced photo mosaic of the deployment area using a ship-towed camera sled and extract relative node positions at centimeter-scale precision. We then compare three established localization algorithms, a centralized least-squares solver and two distributed approaches based on the extended Kalman filter and belief propagation, against this reference. All algorithms achieve relative accuracy below 2 m. The centralized least-squares approach outperforms the distributed approaches by more than 1 m. This demonstrates that the benchmark is sensitive enough to evaluate the performance of different algorithms, offering a path toward noncircular evaluation of underwater localization systems.Publicationtype: Conference Paper11 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Soft-information-based underwater tracer localization with USBL hydrophone arraysLocalization of underwater objects is required for numerous applications, including environmental monitoring, ropeless fishing, and equipment tracking. Ease of deployment and small, low-cost tracers running for years on a battery are incompatible with state-of-the-art solutions, especially in highly challenging multipath environments. We present a passive, soft-information-based localization method leveraging receiver mobility to estimate the 2D-location of a beacon signal with Angle of Arrival observations. The method is shown to be more robust than approaches that make hard decisions on time of arrival observations early on. Furthermore, we present the results in a real-world harbor environment, showing that median localization accuracy below 5m can be achieved.Publicationtype: Conference Paper5 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Resilience against soft faults through adaptivity in spectral deferred correction(Springer, 2026-08-06); ; ; ; As supercomputers grow in hardware complexity, their susceptibility to faults increases, and measures need to be taken to ensure the correctness of results. Some numerical algorithms have certain characteristics that allow them to recover from some types of faults. It has been demonstrated that adaptive Runge–Kutta methods provide resilience against transient faults without adding computational cost. Using recent advances in adaptive step size selection for spectral deferred correction (SDC), an iterative numerical time stepping scheme that can achieve arbitrary order, we show that adaptive SDC can also detect and correct transient faults. Its performance is found to be comparable to that of the dedicated resilience strategy Hot Rod.Publicationtype: Conference Paper41 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Fourier neural operators for Rayleigh–Bénard convection(Springer, 2026-06); ; ; ; We propose an improved Fourier Neural Operator (FNO) for modeling two-dimensional Rayleigh–Bénard convection by predict-ing time increments instead of full solutions, achieving higher accuracy than a standard FNO baseline. The resulting model is compact (314k parameters, 1.26 MB) and fast (7 ms inference), while maintainingsim-ilaraccuracyasdemonstratedinpreviousbenchmarks.WeshowthatalthoughFNOsgeneralizetofinermeshes,accuracyremainslimitedbytheresolutionofthetrainingdata.Publicationtype: Conference Paper45 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Beyond silicon: materials, mechanisms, and methods for physical neural computing(IEEE, 2026-05-08); ; ; ; Physical implementations of neural computation now extend far beyond silicon hardware, encompassing substrates such as memristive devices, photonic circuits, mechanical metamaterials, microfluidic networks, chemical reaction systems, and living neural tissue. By exploiting intrinsic physical processes, such as charge transport, wave interference, elastic deformation, mass transport, and biochemical regulation, these substrates can realize neural inference and adaptation directly in matter. As silicon GPU-centered AI faces growing energy and data-movement constraints, physical neural computation becomes increasingly relevant as a complementary path beyond conventional digital accelerators. This trend is driven in particular by pervasive intelligence, i.e., the deployment of on-device and edge AI across large numbers of resource-constrained systems. In such settings, co-locating computation with sensing and memory can reduce data shuttling and improve efficiency. Meanwhile, physical neural approaches have emerged across disparate disciplines, yet progress remains fragmented, with limited shared terminology and few principled ways to compare platforms. This survey unifies the field by mapping neural primitives to substrate-specific mechanisms, analyzing architectural and training paradigms, and identifying key engineering constraints including scalability, precision, programmability, and I/O interfacing overhead. To enable cross-domain comparison, we introduce a first-order benchmarking scheme based on standardized static and dynamic tasks and physically interpretable performance dimensions. We show that no single substrate dominates across the considered dimensions; instead, physical neural systems occupy complementary operating regimes, enabling applications ranging from ultrafast signal processing and in-memory inference to embodied control and in-sample biochemical decision making.Publicationtype: Journal Article164 29 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , ShipNN: on-device, ship identification from underwater noise(Association for Computing Machinery (ACM), 2026-03); ; ; ; Real-time analysis of underwater acoustic signals is essential for maritime security, environmental monitoring, and wildlife protection. In many real-world deployments, such as remote sensing buoys, autonomous underwater vehicles, and seabed monitoring stations, data must be processed directly on the device due to limited connectivity and energy constraints. These platforms are, for example, tasked with monitoring vessel activity in ecologically sensitive zones, where timely and localized analysis is essential to detect unauthorized presence, enforce protection boundaries, and mitigate acoustic disturbances. Existing approaches mainly focus on coarse ship-type classification or perform individual ship identification at a very limited scale, often restricted to just a handful of vessels. Further, they typically rely on large cloud-based DNNs making them unsuitable for deployment on constrained devices.To close this gap, we present ShipNN, a resource-efficient DNN model for real-time ship identification from underwater acoustic signals, deployable on resource-constrained microcontrollers. ShipNN achieves 94.1% accuracy in individual ship identification, outperforming, for example, ResNet, a much larger baseline model with merely 92.2% accuracy. ShipNN has 32× fewer parameters, 17× less RAM, and 66× less flash storage than this baseline, demonstrating efficient, real-time operation on off-the-shelf microcontrollers.Publicationtype: Conference Paper79 19 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Online passive localization for underwater swarm vehicles using a low-cost acoustic modemWe present an online underwater localization method that enables passive vehicles to localize themselves by overhearing two-way ranging (TWR) communication between an active vehicle and multiple acoustic anchors. Swarm simulations confirm the scalability of the system and show faster position updates, as well as improved accuracy compared to the active benchmark. A field experiment in a shallow water canal using the low-cost BlueROV2 underwater vehicle and ahoi modems has demonstrated the real-world applicability of the system, achieving an RMSE of approximately 0.82m.Publicationtype: Conference Paper132 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , ASV-aided AUV navigation: A field study on nonlinear estimation for localization of low-cost, scalable systems(IEEE, 2025-11-27); ; ; ; Benjamin, MichaelThis work investigates the use of multiple Autonomous Surface Vehicles (ASVs) as Communication/Navigation Aids (CNAs) to enhance the navigation and state estimation of an Autonomous Underwater Vehicle (AUV). Our approach builds on recent advancements in low-cost sensors and platforms, which enable novel AUV applications across fundamental science, commercial industries, and defense. We consider six different combinations of Kalman Filter and Factor Graph localization solutions on three datasets, covering 53 minutes and 3.1 kilometers of operation. We first present the solution using the measurements from all three ASVs, before occluding measurements from two of the ASVs to assess the effect of reduced observability on localization performance.Publicationtype: Conference Paper54 46