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

Contrastive learning of sentence representations has achieved great improvements in several natural language processing tasks. However, the supervised contrastive learning model trained on the natural language inference (NLI) dataset is insufficient to elucidate the semantics of sentences since it is prone to make a prediction based on heuristics. Herein, by using the ParsEVAL and the word overlap metric, it is shown that sentence pairs in the NLI dataset have strong syntactic similarity and propose a framework to compensate for this problem in two aspects. 1) Apply simple syntactic transformations to the hypothesis and 2) expand the objective to SupCon Loss to leverage variants of sentences. The method is evaluated on semantic textual similarity (STS) tasks and transfer tasks. The proposed methods improve the performance of the BERT-based baseline in STS Benchmark and SICK Relatedness by 1.48% and 2.2%. Furthermore, the model achieves 82.65% on the HANS benchmark dataset, to the best of our knowledge, which is a state-of-the-art performance demonstrating that our approach is effective in grasping semantics without heuristics in the NLI dataset at supervised contrastive learning. The code is available at https://github.com/whnhch/Break-the-Similarity.

Soft robots exhibit significant flexibility but normally lack stability owing to their inherent low stiffness. Current solutions for achieving variable stiffness or implementing lock mechanisms tend to involve complex structures. Additionally, passive solutions like bistable and multistate mechanisms lack spatial stable characteristics. This study presents a novel shape memory alloy (SMA) modular robot with spatially stable structure, by utilizing gooseneck as the backbone. This is the first time that a concept of spatially stable structure is proposed. When the power is off, the robot can still maintain its current posture in three-dimensional space and resist external disturbance. The SMA spring and gooseneck are characterized, elucidating the mechanism behind achieving spatial stability. Then, a controller based on the inverse kinematics is designed, and validated by experiments. The results demonstrate the structural stability of the robot. Specifically, it can withstand a maximum external force of 2.5 N (0.0875 Nm) when bent at an angle of 20° without consuming energy. Moreover, with the assistance of the SMA spring, this resistance capacity surpasses 5 N (0.175 Nm).

In nature, gliding birds frequently execute intricate flight maneuvers such as aerial somersaults, perched landings, and swift descents, enabling them to navigate obstacles or hunt prey. It is evident that birds rely on different wing–tail configurations to accomplish a wide range of aerial maneuvers. For traditional fixed-wing unmanned aerial vehicles (UAVs), pitch control primarily comes from the tail's elevators, while adjusting flight lift and drag involves deploying wing flaps. Although these designs ensure reliable flight, they compromise the drones’ maneuverability to maintain longitudinal stability. Therefore, the study introduces a biomimetic morphing wing UAV, and presents a pitch control strategy that simultaneously engages morphing wings, ailerons, and tail elevators. The pull-up maneuver tests indicate that the proposed control method results in a pitch rate that is approximately 2.5 times greater than when using only the elevator control. A closed-loop control system for the drone is also established. The closed-loop flight experiment, which tracks a 45° pitch angle, demonstrates the effectiveness of the proposed coupled control method in adjusting the flight attitude. In addition, during cruising, the UAV employs three configurations, straight wing, forward-swept wing, and back-swept wing, to cater to different mission objectives and augment its flight capabilities.

Soft robots can adaptively interact with unstructured environments. However, nonlinear soft material properties challenge modeling and control. Learning-based controllers that leverage efficient mechanical models are promising for solving complex interaction tasks. This article develops a closed-loop pose/force controller for a dexterous soft manipulator enabling dynamic pushing tasks using deep reinforcement learning. Force tests investigate the mechanical properties of a soft robot module, resulting in orthogonal forces of $��9�-�13�$ N. Then, the policy is trained in simulation leveraging a dynamic Cosserat rod model of the soft robot. Domain randomization mitigate the sim-to-real gap while careful reward engineering induced pose and force control even without explicit force inputs. Despite the approximate simulation, the sim-to-real transfer achieved an average reaching distance of $��34�\pm �14�$ mm ($��8.1�\%�L�\pm �3.4�\%�L�$), an average orientation error of $��0.40�\pm �0.29�$ rad ($���23��°�\pm ��17��°�$) and applied pushing forces up to $�3$ N. Such performance is reasonable for the intended assistive tasks of the manipulator. The exper

Histopathology plays a pivotal role in medical diagnostics. In contrast to preparing permanent sections for histopathology, a time-consuming process, preparing frozen sections is significantly faster and can be performed during surgery, where the sample scanning time should be optimized. Super-resolution techniques allow imaging of histopathalogical samples in lower magnification, thus sparing scanning time. Herein, a new approach is presented to super-resolution of histopathological frozen sections, with focus on achieving better distortion measures, rather than pursuing photorealistic images that may compromise critical diagnostic information. Our deep-learning architecture focuses on learning the error between interpolated images and real images; thereby generating high-resolution images while preserving critical image details, which reduces the risk of diagnostic misinterpretation. This is done by leveraging the loss functions in the frequency domain and assigning higher weights to the reconstruction of complex, high-frequency components. In comparison with existing methods, significant improvements are obtained in terms of distortion metrics, improving the pathologist's clinical decisions. This approach has a great potential to provide faster frozen-section imaging, with less scanning, speeding up intraoperative decisions, while preserving the high-resolution details in the imaged sample.

Leveraging the unique attributes of functional soft materials to generate force and deformation, significant advancements in soft actuators are driving the evolution of smart robotics. Liquid metals (LMs), known for their high deformability and tunable morphology, demonstrate remarkable actuating capabilities through controllable surface tension. Inspired by the predation method of chameleons, this work introduces a bioinspired LM actuator (BLMA) by modulating the morphology of LM. This BLMA enables high-strain (up to 170%) actuation by precisely directing LM droplets toward an electrode. Various parameters affecting the BLMA's actuating performance are explored. Notably, the application of a reductive voltage induces rapid solidification of supercooled LM, facilitating phase transition at room temperature. The solidified LM enhances its holding force of BLMA by over 1000 times. To underscore the superior capabilities of the BLMA, diverse applications, such as a complex two-dimensional plane actuator, a stepper motor with adjustable step intervals, a phase transition-controlled relay, and a laser code lock actuation gate set, are presented. It is anticipated that the exceptional characteristics of the BLMA will propel advancements in the realms of soft robotics and mechatronics.

Monolithic three-dimensional (M3D) integration advances integrated circuits by enhancing density and energy efficiency. Ferroelectric thin-film transistors (Fe-TFTs) attract attention for neuromorphic computing and back-end-of-the-line (BEOL) compatibility. However, M3D faces challenges like increased runtime temperatures due to limited heat dissipation, impacting system reliability. This work demonstrates the effect of temperature impact on single-gate (SG) Fe-TFT reliability. SG Fe-TFTs have limitations such as read-disturbance and small memory windows, constraining their use. To mitigate these, dual-gate (DG) Fe-TFTs are modeled using technology computer-aided design, comparing their performance. Compute-in-memory (CIM) architectures with SG and DG Fe-TFTs are investigated for deep neural networks (DNN) accelerators, revealing heat's detrimental effect on reliability and inference accuracy. DG Fe-TFTs exhibit about 4.6x higher throughput than SG Fe-TFTs. Additionally, thermal effects within the simulated M3D architecture are analyzed, noting reduced DNN accuracy to 81.11% and 67.85% for SG and DG Fe-TFTs, respectively. Furthermore, various cooling methods and their impact on CIM system temperature are demonstrated, offering insights for efficient thermal management strategies.

The production industry is challenged to become more flexible and efficient while coping with a variety of disruptive events, such as natural disasters, infrastructure blockages, or economic crises. From the individual station on a production line to the global supply chain, everything is connected, making regulation and control a complex task. Biological molecular processes, such as the metabolism of living organisms or the cell cycle, are also extremely complex processes that can be compared to industrial production processes, both of which involve a series of intermediate steps and products. Thanks to (self-)regulatory mechanisms that have evolved over time, these biological mechanisms are very efficient and robust in the face of perturbations. This article proposes an explanatory representation of these complex processes, considering both biological and technical aspects. The aim is to facilitate biomimetic transfer of biological regulation mechanisms into the technical domain. It presents concepts for biomimetic regulation of production lines and sourcing strategies and introduces a workflow for generating digital twins. This workflow is inspired by the cell cycle checkpoints, which ensure that only perfect copies of DNA are passed on during cell replication. By leveraging this understanding, the production industry can potentially improve its own processes and efficiency.

Reservoir computing (RC) possesses a simple architecture and high energy efficiency for time-series data analysis through machine learning algorithms. To date, RC has evolved into several innovative variants. The next generation reservoir computing (NGRC) variant, founded on nonlinear vector autoregression (NVAR) distinguishes itself due to its fewer hyperparameters and independence from physical random connection matrices, while yielding comparable results. However, NGRC networks struggle with massive Kronecker product calculations and matrix-vector multiplications within the read out layer, leading to substantial efficiency challenges for traditional von Neumann architectures. In this work, a hybrid digital-analog hardware system tailored for NGRC is developed. The digital part is a Kronecker product calculation unit with data filtering, which realizes transformation of nonlinear vector of the input linear vector. For matrix-vector multiplication, a computing-in-memory architecture based on resistive random access memory array offers an energy-efficient hardware solution, which markedly reduces data transfer and greatly improve computational parallelism and energy efficiency. The predictive capabilities of this hybrid NGRC system are validated through the Lorenz63 model, achieving a normalized root mean square error (NRMSE) of 0.00098 and an energy efficiency of 19.42TOPS W⁻¹.

The precise control of soft and continuum robots requires knowledge of their shape, which has, in contrast to classical rigid robots, infinite degrees of freedom. To partially reconstruct the shape, proprioceptive techniques use built-in sensors, resulting in inaccurate results and increased fabrication complexity. Exteroceptive methods so far rely on expensive tracking systems with reflective markers placed on all components, which are infeasible for deformable robots interacting with the environment due to marker occlusion and damage. Here, a regression approach is presented for three-dimensional key point estimation using a convolutional neural network. The proposed approach uses data-driven supervised learning and is capable of online markerless estimation during inference. Two images of a robotic system are captured simultaneously at 25 Hz from different perspectives and fed to the network, which returns for each pair the parameterized key point or piecewise constant curvature shape representations. The proposed approach outperforms markerless state-of-the-art methods by a maximum of 4.5% in estimation accuracy while being more robust and requiring no prior knowledge of the shape. Online evaluations on two types of soft robotic arms and a soft robotic fish demonstrate the method's accuracy and versatility on highly deformable systems.