Advanced intelligent systems (Weinheim an der Bergstrasse, Germany)最新文献

Symbolic Reservoir Computing Using Memristive Cellular Automata

In quest of a neuro-symbolic system with the strengths of both neural network and symbolic approaches, Cheng Ma, Wenrui Duan, Huanglong Li, and co-workers have proposed and experimentally demonstrated a memristor crossbar array-based cellular automata (symbolic) for reservoir computing (neural network), exhibiting high parallelism, high energy efficiency and high robustness (due to hyper-dimensionality) against inherent variations of the emerging hardware. More details can be found in the Research Article DOI: 10.1002/aisy.202500241.

Pneumatic Actuators

A wearable device using pneumatic actuators and origami pumps was proposed to generate multiple feedbacks including normal force, shear force, and vibration. The actuators and origami pumps are made of soft materials to reduce weight significantly, thereby increasing wearability while delivering appropriate force and air pressure. More details can be found in the Research Article by Youngsu Cha and co-workers (DOI: 10.1002/aisy.202500374).

Microrobots

Magnetic microrobots hold a promising future in many industries but are challenging to fabricate. An automated fabrication process is vital for accurate, reproducible products that can be developed quickly and efficiently. Grippers are one of the most widely used microrobots capable of numerous tasks such as grasping, crawling and targeted delivery. This cover is designed to represent the automated fabrication of magnetic microdevices. More details can be found in the Research Article by Onaizah Onaizah and co-workers (DOI: 10.1002/aisy.202500051).

Smart Prosthetic Sockets

Residual limb volume fluctuations can significantly compromise prosthetic socket fit, causing discomfort and instability, and often requiring frequent adjustments or replacements. This study presents a transfemoral socket with a motorized cable-driven mechanism and sensorized liner, adapting to volume changes via interface pressure monitoring. Controlled in open- or closed-loop modes through a custom mobile app, it enhances comfort, stability, and user autonomy. More details can be found in the Research Article by Linda Paternò and co-workers (DOI: 10.1002/aisy.202400995).

Magnetically Actuated Artificial Cilia

A magnetically actuated artificial cilia (MAAC) platform is developed for reversible microflow switching and particle manipulation. By modulating the beating frequencies via external magnetic fields, the system enables shear-driven particle trapping, directional release, and in situ microfluidic mixing. This strategy offers a promising, adaptive and versatile alternative for superior microscale flow control in robotics, biological processing, and lab-on-a-chip technologies. More details can be found in the Research Article by Chia-Yuan Chen and co-workers (DOI: 10.1002/aisy.202500431).

Soft Electronics

A Lorentz-force-driven, liquid metal–embedded surface delivers rapid, reversible 3D shape morphing for precise low-velocity flow control. With minimal power, it modulates near-wall flows in real time, offering versatile, programmable actuation for small UAVs, bio-inspired aerodynamics, and environmental sensing—bridging soft electronics with advanced fluid dynamics. More details can be found in the Research Article by Donghyun You, Leonardo P. Chamorro, Xinchen Ni, John A. Rogers, and co-workers (DOI: 10.1002/aisy.202500457).

Soft Wearable Glove

A soft wearable modular assistive glove, driven by miniature foldable pouch motor unit (MFPMU), is developed for hand rehabilitation and daily assistance. The MFPMU combine bending and elongation to naturally adapt to finger joint motion, achieving a fingertip force of 2.34 N at 50 kPa. The actuators generate a maximum torque of 240 mN·m and can be replaced within 30 seconds, offering high adaptability and user convenience. More details can be found in the Research Article by Jianbin Liu and co-workers (Doi: 10.1002/aisy.202500274).

Drying droplets of complex biofluids reveal a rich interplay of evaporation-driven flows, phase segregation, and self-assembly, resulting in intricate patterns that encode significant spatio-temporal information. Whereas prior studies have predominantly emphasized spatial analysis, this work advances a framework that incorporates both spatial and temporal dimensions, leveraging machine learning (ML) for accurate compositional classification of blood samples. Systematic variations in initial concentration manifest as quantifiable differences in drying behavior, captured through spatio-temporal imaging. Statistical features extracted from these image sequences enable traditional MLs to achieve 99% classification accuracy, outperforming deep learning (DL) that achieves 96% accuracy when tested directly on new image data. Gradient-weighted Class Activation Mapping (Grad-CAM) indicates that DL focuses on highly localized textural regions, revealing that the dynamic evolution of drying encodes more information than static end-point images suggest. Importantly, the proposed framework extends beyond blood diagnostics, demonstrating broad applicability to microbial suspensions, protein solutions, and liquid crystals. This work introduces an interpretable, low-volume, and sustainable analytical method, establishing drying droplets as a powerful, high-throughput readout for fluid behavior across scientific disciplines.

Many organisms like earthworms with soft bodies and simple nervous systems can sense and respond to stimuli, conducting complex tasks such as navigation, foraging, and transporting objects. However, most soft robots currently require rigid semiconductor-based electronics for sensing and control, limiting the benefits of their soft bodies and posing challenges for integration. To address these limitations, a stimuli-responsive electrofluidic nervous system (SENS) composed of soft materials to realize signal generation, multimodal stimuli-sensing, and decision-making for multiactuator soft electroactive robots is proposed. SENS is composed of multiple fluidic switches, which are driven by electroactive actuators and by external stimuli such as force and heat transduced into fluidic movement by sensing receptors. Electrofluidic circuits are created using these switches to achieve self-starting oscillating circuits that control input voltages to actuators and mode-selection units that activate specific oscillating circuits based on applied external stimuli to achieve stimuli-responsive behaviors. Utilizing SENS, a soft crawling robot that can change its direction of motion in response to tactile and heat stimuli is realized. The robot is made of a dielectric elastomer actuator and two electroadhesion actuators. Furthermore, an untethered soft robot has been developed with a miniaturized SENS and an onboard constant voltage power source, which can exhibit unidirectional motion. This work constitutes a step toward developing electronics-free, entirely soft autonomous robots capable of versatile and adaptive tasks.

The increasing demands of data-centric applications continue to expose the fundamental bandwidth bottleneck of the von Neumann architecture, where memory and computation are separated. Analog processing-in-memory (PIM) offers a promising pathway to overcome this limitation, and charge-trap flash (CTF) synapses stand out as attractive candidates owing to their scalability, multilevel storage, and complementary metal-oxide-semiconductor (CMOS) compatibility. In this work, a hardware-level analog PIM system is presented that integrates CTF synapse arrays, CMOS-based wordline drivers, and a novel successive integration-and-rescaling (SIR) neuron circuit in a chip-on-board configuration. Unlike conventional neuron designs that rely heavily on analog-to-digital conversion, the proposed SIR neuron performs bit-sliced accumulation entirely in the analog domain. This architecture not only minimizes analog-to-digital converter overhead but also achieves excellent linearity through input-node stabilization and functional capacitor separation, thereby enhancing both computational accuracy and area efficiency. The fabricated system is validated through handwritten digit classification on the modified national institute of standards and technology dataset, achieving an accuracy of 72.93%, which is only 3.91 percentage points lower than the software baseline under identical precision. These results underscore the pivotal role of the SIR neuron in bridging device-level innovations with system-level integration, positioning CTF-based analog PIM as a scalable and energy-efficient platform for neuromorphic computing.