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

In this study, it is aimed to establish a novel method based on a deep-learning-guided surface-enhanced Raman spectroscopy (SERS) technique to achieve rapid and accurate classification of vaginal cleanliness levels. We proposed a variational autoencoder (VAE) approach to enhance spectral quality, coupled with a deep learning algorithm long short-term memory (LSTM) neural network to analyze SERS spectra produced by vaginal secretions. The performance of various machine learning (ML) algorithms is assessed using multiple evaluation metrics. Finally, the reliability of the optimal model is tested using blind test data (N = 10/group for each cleanliness level). The data quality of the SERS fingerprints of four types of vaginal secretions is significantly improved after VAE decoding and reconstruction. The signal-to-noise ratio of the generated spectra increased from the original 2.58–11.13. Among all algorithms, the VAE–LSTM algorithm demonstrates the best prediction ability and time efficiency. Additionally, blind test datasets yielded an overall accuracy of 85%. In this study, it is concluded that the deep-learning-guided SERS technique holds significant potential in rapidly distinguishing between different levels of vaginal cleanliness through human vaginal secretion samples. This contributes to the efficient diagnosis of vaginal cleanliness levels in clinical settings.

Conventional continuum robots have outstanding flexibility and dexterity. However, when the robot needs to interact with the environment, the softness may affect the performance of the robot. Especially in transport tasks, the softness of continuum robots can lead to handling failures and drastic drops in precision. The variable stiffness continuum robot combines the advantages of flexibility and rigidity, which is conducive to expanding the application scenarios of flexible continuum robots. This article proposes a flexible continuum robot that simultaneously realizes variable stiffness, shape-aware, and self-heating functions using liquid metal. The low-temperature phase transition property of liquid metal is utilized to realize the variable stiffness function; the overall stiffness of the robot can reach the range of 18.5–183 N m⁻¹, which can realize a tenfold stiffness gain. The conductivity of liquid metal is utilized to develop the shape-aware function, and the monitoring accuracy is within 5%. At the same time, this article utilizes the liquid metal's resistive thermal effect to realize heating function, so that the robot no longer needs heating systems such as heating wires and can realize the phase transition by energizing itself. Based on this design, the robot arm can realize the transition between maximum and minimum stiffness within 240 s.

Surface functional microstructures exhibit extensive application requirements in an array of breakthrough areas. One critical problem that restricts their industrial application is the lack of scalable fabrication techniques due to the limitation of conventional machine tools. This study proposes a scalable surface texturing technique using a portable small (30 × 19 × 22 mm) three-leg robot that walks and works on the workpiece surface. Due to the elliptical tool vibration, microgrooves can be created on the workpiece surface periodically; meanwhile, the machining force drives the robot to walk forward. Surface texturing experiments are conducted on aluminum and copper workpieces to explore the machining performance of the small robot. The robot can reach a maximum moving velocity of 6.3 mm s⁻¹ and can produce microstructures with a spacing of 4–14 μm on workpiece surfaces. Owing to its unique working principle, the small robot can maintain a constant depth of cut, demonstrating its capacity to adapt to the surface waviness of the workpiece. Finally, the motion straightness of the robot is greatly improved by combining it with the auxiliary track, and multiline microstructures are obtained. In short, the developed small robot presents a promising solution to the challenge of scalable surface texturing.

The lack of a sufficient and efficient way to simultaneously perceive general underwater mechanical stimuli, physical contact, and fluidic flow has been a bottleneck for many aquatic applications. To address this challenge, dynamics-oriented underwater mechanoreceptor interface (DOUMI), a bioinspired mechanoreception system that realizes simultaneous contact and flow perception using a single receptor, is introduced. This receptor, response-elevated-and-expanded hair-like tactile mechanoreceptor (REEM), is inspired by the mechanoreceptive mechanism of aquatic arthropods. REEM combines structural features from different mechanoreceptive sensilla, enabling it to capture a wide range of stimulus dynamics. Under different stimuli, REEM encodes stimuli dynamics as its oscillations with distinct spectral attributes. Those oscillations are efficiently transferred through mechanical processes and imaging, enabling vision-based extraction and further analysis. Therefore, by evaluating the oscillation dynamics with tailored wavelet-based indices, DOUMI can distinguish between contact- and flow-induced oscillations at each receptor unit with 90.5% accuracy. Furthermore, DOUMI provides comprehensive 2D mechanoreception with a scalable array of REEMs, delivering capabilities like stimuli spatiotemporal visualization, flow trend detection, and scenario classification with an accuracy of 99.5%. With its robustness and operational efficiency in underwater environments, DOUMI can be easily adapted to existing applications using common materials and hardware, establishing a new, streamlined paradigm for underwater general mechanoreception.

Solar cell research aims to improve power conversion efficiency (PCE). This field has an extensive body of literature on the Web of Science. For researchers, it is impossible to understand the development of the entire field comprehensively through traditional reading methods. Knowledge is recorded in the literature by text and numbers. Researchers acquire knowledge through literature surveying, text reading, and thinking. The conversion from text and numbers to knowledge can be automatically completed by machines, which can avoid path-dependent perspectives. In this work, an intelligent machine learning method for literature structure delineation and information extraction is proposed. As an example, a knowledge base of organic solar cells (OSCs) is extracted including topic analysis of literature, numerical characteristics of performance, and material information. Seven major research directions of OSCs are identified. The correlations between key performance parameters, including PCE, short-circuit current density (J_SC), open-circuit voltage (V_OC), and fill factor (FF), are revealed from text mining. A donor–acceptor material map of PCE is constructed which provides a road map for OSCs, indicating the bottleneck of this field. Moreover, the method of machine intelligence developed here can be used in any other materials field, aiding a comprehensive understanding of the development quickly.

Reconstructing Soft Robotic Touch via In-Finger Vision

The research by Fang Wan, Chaoyang Song, and co-workers (see article number 2400022) introduces a vision-based approach for learning proprioceptive interactions using Soft Robotic Metamaterials (SRMs). By reconstructing shape and touch during physical engagements, the authors achieve real-time, precise estimations of the soft finger mesh deformation in virtual environments. This innovation enhances the adaptability in 3D interactions and suggests promising applications in human–robot collaboration and touch-based digital twin interactions, bridging the gap between physical and virtual worlds via a multi-modal soft touch.

Cable-Actuated Soft Manipulator Based on Deep Reinforcement Learning

In article number 2400112, Juntian Qu and co-workers propose a type of modified TD3 (twin delayed deep deterministic policy gradient) algorithm in combination with LSTM (long short-term memory) neural networks to control the cable-driven soft manipulator. Multi-scenario and multi-task experiments are carried out based on the soft manipulator, such as precisely placing a 6 mm diameter ball into a 10 mm diameter glass bottle and accurately retrieving a shell from within an L-shaped pipe using the soft manipulator.

Reprogrammable, Recyclable Origami Robots

The research highlighted by this cover focuses on creating innovative paper-based origami robots that transform a simple 2D sheet into complicated 3D shapes using magnetic programming (see article number 2400082). Sang Min Won, Yoonseok Park, and co-workers embed these biodegradable robots with conductive nanoparticles and electrical components, enabling them to monitor environmental conditions and repair complex machinery. The authors believe that these advancements will broaden the use of origami robots in various areas of soft robotics, offering versatile, eco-friendly solutions.

Liquid Metal Chameleon Tongues – Bioinspired Soft Actuators

Article number 2400231 by Hongda Lu, Shi-Yang Tang, Weihua Li, and co-workers presents a bio-inspired liquid metal soft actuator inspired by the predation behavior of chameleons. By delicately modulating its surface tension and phase transition, the actuator achieves reciprocating motion, exhibiting high strain rate, enhanced adhesive force, and reconfigurability. These superior performances highlight a potential in cargo delivery, complex 2D motion, and advanced smart mechatronics.

Automatic design is an appealing approach to realizing robot swarms. In this approach, a designer specifies a mission that the swarm must perform, and an optimization algorithm searches for the control software that enables the robots to perform the given mission. Traditionally, research in automatic design has focused on missions specified by a single design criterion, adopting methods based on single-objective optimization algorithms. In this study, we investigate whether existing methods can be adapted to address missions specified by concurrent design criteria. We focus on the bi-criteria case. We conduct experiments with a swarm of e-puck robots that must perform sequences of two missions: each mission in the sequence is an independent design criterion that the automatic method must handle during the optimization process. We consider modular and neuroevolutionary methods that aggregate concurrent criteria via the weighted sum, hypervolume, or $��l^2�$-norm. We compare their performance with that of Mandarina, an original automatic modular design method. Mandarina integrates Iterated F-race as an optimization algorithm to conduct the design process without aggregating the design criteria. Results from realistic simulations and demonstrations with physical robots show that the best results are obtained with modular methods and when the design criteria are not aggregated.