Repository for Publications and Research Data
News from the ETH Library
Recently Added
Odermatt, Ingrid A. (2026)
Photonically linked three-dimensional neural networks based on memristive blinking neurons
Item type: Journal Article
Zhou, Yue; Fang, Yuetong; Gisler, Raphael; et al. (2026)
The continuing development of artificial intelligence requires more powerful computing architectures. However, the large footprint of complementary-metal-oxide-semiconductor-based neurons and constraints on electric routing hinder the scaling of conventional artificial neurons and their synaptic connectivity. Here we show that memristive blinking neurons can be used to build scalable photonically linked three-dimensional neural networks. Our artificial neuron is based on a silver/poly(methyl methacrylate)/silver metal-insulator-metal memristive switching in-plane junction. Its resistive switching relies on atomic-scale filamentary dynamics and the device emits photon pulses on integrating a critical number of incoming electrical spikes, which eliminates the need for bulky peripheral circuit read-out and electrical wiring for transmitting signals. We use the memristive blinking neuron, which has a footprint of 170 nm x 240 nm, to build a photonically linked three-dimensional spiking neural network. We show that the network can perform a four-class classification task within the Google Speech dataset with an accuracy of 91.51%. We also create a high-density artificial neuron array with a pitch of 1 mu m and show that it can perform an MNIST classification task with an accuracy of 92.27%.
Deep-Learning-Assisted Kalman Filter for GNSS/MEMS IMU Integration in GNSS-Denied Environments
Item type: Journal Article
Li S.; Mikhaylov M.; Pany T.; et al. (2026)
This study introduces a novel deep-learning-assisted Kalman filter (DLKF) for integrating global navigation satellite system (GNSS) and inertial navigation system (INS) measurements, aiming to improve navigation accuracy and robustness in GNSS-denied environments. GNSS/INS integration is critical for autonomous vehicles to maintain continuous navigation when GNSS signals are obstructed. Traditional model-based Kalman filters (MBKFs), while widely employed for GNSS/INS integration, face challenges such as system model deficiencies, incorrect noise assumptions, and complex inertial measurement unit (IMU) error characteristics, especially for micro-electromechanical system IMUs. To address these challenges, the DLKF integrates a deep neural network (DNN) with the Kalman filter, leveraging its ability to learn non-linear system dynamics and noise statistics to provide adaptive Kalman gain and accurate IMU error estimation. Our approach further addresses practical challenges such as training with asynchronous sensor data and adapting to an error-state MBKF structure. This method also leverages DNN capabilities while retaining the interpretability of model-based filtering. In addition, we design specialized training strategies to enhance model stability and performance, particularly under GNSS outage conditions. The proposed method was evaluated on simulated three-dimensional data sets and real-world scenarios. The results demonstrate that the DLKF achieves improved navigation accuracy and reduced position errors during GNSS outages compared with the MBKF. More importantly, the proposed approach provides a consistent and interpretable framework for integrating DNNs into the well-established MBKF-based GNSS/ INS integration framework. These findings highlight the potential of combining data-driven methods with traditional model-based approaches to address the limitations of GNSS/INS integration and to provide enhanced solutions for autonomous systems in challenging environments.
In-space manufacturing of optical lenses: Fluidic Shaping aboard the International Space Station
Item type: Journal Article
Luria O.; Elgarisi M.; Stibbe E.; et al. (2026)
In-space manufacturing technologies are considered vital for enabling advanced space missions and addressing logistical limitations of space exploration. While additive manufacturing has progressed rapidly, it still falls short of delivering the ultra-smooth surfaces required for optical elements. Fluidic Shaping is a novel method that harnesses surface tension under microgravity to form optical components with exceptionally smooth surfaces. This study demonstrates the feasibility and potential of Fluidic Shaping as a method for manufacturing optical components in space through two experiments performed aboard the International Space Station (ISS) during the Ax-1 mission. The first experiment involved fabricating centimeter-scale polymer lenses, solidifying them via ultraviolet (UV) curing, and analyzing the resultant optics upon their return to Earth. While sub-nanometric surface smoothness was achieved, some polymer lenses displayed unexpected thermo-chemical deformations, indicating complex polymerization dynamics unique to the microgravity environment. In the second experiment, a 172 mm diameter water lens was deployed, confirming Fluidic Shaping’s scalability and demonstrating basic optical functionality through image analysis. These experiments collectively underline the technique’s relevance for both small-scale optics and large-aperture applications. Our results highlight critical considerations for future research, including optimizing polymerization processes and refining liquid-handling methods to advance practical, in-space optical manufacturing capabilities.
Item type: Person