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

Kirigami Flexible Grippers

A kirigami multi-stable flexible gripper with trigger structure is presented by Xiuting Sun and co-workers in article number 2400038. The advanced trigger structure is designed to make the flexible gripper enable an energy-free switching behavior between deployed and curled configurations in symmetrical and asymmetrical planes. The proposed gripper is appropriate for the capture of space debris and demonstrates significant capture capability for moving targets.

Biomimetic Synthesis

In article 2300467, Weiqi Fu and co-workers use machine learning techniques to design peptides with silicifying functionality. Inspired by the exquisite nanosilica structures from nature, a deep learning model, based on antimicrobial peptide migration learning, is developed with the inputs of a comprehensive collection of silicifying peptides from diatoms to achieve the biomimetic synthesis of nanosilica. The newly designed silicified peptides could facilitate the development of new biosensors and drug delivery systems. [Image by Jiwei Chen and Mengsheng Xia.]

Targeted Mass Spectrometry Data Analysis

The application of liquid chromatography-tandem mass spectrometry (LC-MS/MS) has facilitated the earlier detection and diagnosis of diseases preceding the manifestation of symptoms, but data analysis is complicated for clinical application. Integrating an automated machine learning pipeline can optimize LC-MS/MS data processing and analysis, even with limited training datasets. Machine learning pipelines can also implement an active learning nested model to mitigate bias from imbalanced training datasets, providing more accurate clinical proteomic analysis and disease diagnostic results. For more details, refer to article number 2300773 by Jia Fan, Duran Bao, and co-workers.

We offer a new in-depth investigation of global path planning (GPP) for unmanned ground vehicles, an autonomous mining sampling robot named ROMIE. GPP is essential for ROMIE's optimal performance, which is translated into solving the traveling salesman problem, a complex graph theory challenge that is crucial for determining the most effective route to cover all sampling locations in a mining field. This problem is central to enhancing ROMIE's operational efficiency and competitiveness against human labor by optimizing cost and time. The primary aim of this research is to advance GPP by developing, evaluating, and improving a cost-efficient software and web application. We delve into an extensive comparison and analysis of Google operations research (OR)-Tools optimization algorithms. Our study is driven by the goal of applying and testing the limits of OR-Tools capabilities by integrating Reinforcement Learning techniques for the first time. This enables us to compare these methods with OR-Tools, assessing their computational effectiveness and real-world application efficiency. Our analysis seeks to provide insights into the effectiveness and practical application of each technique. Our findings indicate that Q-Learning stands out as the optimal strategy, demonstrating superior efficiency by deviating only 1.2% on average from the optimal solutions across our datasets.

As the third-generation network succeeding artificial neural networks (ANNs), spiking neural networks (SNNs) offer high robustness and low energy consumption. Inspired by biological systems, the limitations of low robustness and high-power consumption in deep reinforcement learning (DRL) are addressed by introducing SNNs. The Brain Q-network (BrainQN) is proposed, which replaces the neurons in the classic Deep Q-learning (DQN) algorithm with SNN neurons. BrainQN is trained using surrogate gradient learning (SGL) and ANN-to-SNN conversion methods. Robustness tests with input noise reveal BrainQN's superior performance, achieving an 82.14% increase in rewards under low noise and 71.74% under high noise compared to DQN. These findings highlight BrainQN's robustness and superior performance in noisy environments, supporting its application in complex scenarios. SGL-trained BrainQN is more robust than ANN-to-SNN conversion under high noise. The differences in network output correlations between noisy and original inputs, along with training algorithm distinctions, explain this phenomenon. BrainQN successfully transitioned from a simulated Pong environment to a ball-catching robot with dynamic vision sensors (DVS). On the neuromorphic chip PAICORE, it shows significant advantages in latency and power consumption compared to Jetson Xavier NX.

Dimension reduction aims to project a high-dimensional dataset into a low-dimensional space. It tries to preserve the topological relationships among the original data points and/or induce clusters. NetDRm, an online dimensionality reduction method based on neural ensemble learning that integrates different dimension reduction methods in a synergistic way, is introduced. NetDRm is designed for datasets of multidimensional points that can be either structured (e.g., images) or unstructured (e.g., point clouds, tabular data). It starts by training a collection of deep residual encoders that learn the embeddings induced by multiple dimension reduction methods applied to the input dataset. Subsequently, a dense neural network integrates the generated encoders by emphasizing topological preservation or cluster induction. Experiments conducted on widely used multidimensional datasets (point-cloud manifolds, image datasets, tabular record datasets) show that the proposed method yields better results in terms of topological preservation ($��R_{�\text{NX}�}�$ curves), cluster induction (V measure), and classification accuracy than the most relevant dimension reduction methods.

Detecting looming signals for collision avoidance encounters challenges in real-world scenarios, where moving backgrounds can interfere as an agent navigates through complex natural environments. Remarkably, even insects with limited neural systems adeptly respond to looming stimuli while in motion at high speeds. Existing insect-inspired looming detection models typically rely on either motion-pathway or feature-pathway signals, yet both are susceptible to dynamic visual scene interference. Coordinating interneuron signals from both pathways can enhance the looming detection performance under dynamic conditions. An artificial neural network is employed to construct a combined-pathway model based on Drosophila anatomy. The model outperforms state-of-the-art bio-inspired looming-detection models in tasks involving dynamic backgrounds, simulated by animated 2D-moving natural scenes or recorded in reality when an unmanned aerial vehicle performs obstacle collision avoidance tasks. Notably, by combining neural anatomy architecture and appropriate multiobjective tasks, the model exhibits convergent neural dynamics with biological counterparts post-training, offering network explanations and mechanistic insights. Specifically, a multiplicative interneuron operation enhances looming signal patterns and reduces background interferences, generalizing to more complex scenarios, such as AirSim 3D environments and real-world situations. The work introduces testable biological hypotheses and a promising bioinspired solution for looming detection in dynamic visual environments.

In response to increasing demand for robotics manipulation, accurate vision-based full pose estimation is essential. While convolutional neural networks-based approaches have been introduced, the quest for higher performance continues, especially for precise robotics manipulation, including in the Agri-robotics domain. This article proposes an improved transformer-based pipeline for full pose estimation, incorporating a Depth Refinement Module. Operating solely on monocular images, the architecture features an innovative Lighter Depth Estimation Network using a Feature Pyramid with an up-sampling method for depth prediction. A Transformer-based Detection Network with additional prediction heads is employed to directly regress object centers and predict the full poses of the target objects. A novel Depth Refinement Module is then utilized alongside the predicted centers, full poses, and depth patches to refine the accuracy of the estimated poses. The performance of this pipeline is extensively compared with other state-of-the-art methods, and the results are analyzed for fruit picking applications. The results demonstrate that the pipeline improves the accuracy of pose estimation to up to 90.79% compared to other methods available in the literature.

This article investigates a reach-avoid game and proposes a cooperative path planning algorithm for a target–pursuers (TP) coalition to capture an evader. In the game, the target aims to bait and escape from the evader, and the pursuer aims to capture the evader. Due to imperfect observations, the TP coalition has uncertain information of the evader's state, while the evader is assumed to have perfect observation. The game model is constructed by formulating the optimization problems for each player in a receding horizon fashion. Then, to counter the evader effectively, the TP coalition constructs a virtual evader using the belief information from a Kalman filter. And a chance constraint optimization problem is constructed to predict the virtual evader's trajectory under uncertainties. The TP coalition can capture the actual evader by generating a robust counter-strategy against the virtual evader with a chance constraint feasible set. Next, to compute the Nash equilibrium of the TP coalition's subjective game, an iterative algorithm is designed that combines the iterative best response and the distributed alternating direction method of multiplier algorithms. Finally, the effectiveness of the algorithm is validated through simulations and experiments.

The current trajectory planning methods for multi-robot systems face challenges due to high computational burden and inadequate adaptability in complex constrained environments, obstructing efficiency improvements in production and logistics. This article presents an innovative solution by integrating model predictive contouring control (MPCC) and long short-term memory (LSTM) networks for real-time trajectory planning of multiple mobile robots. Based on the datasets generated by MPCC, a customized LSTM network is constructed to learn the collaborative planning behavior from these datasets offline, subsequently producing smooth and efficient trajectories online with a low computational burden. Moreover, a hybrid control scheme, incorporating a lidar-based safety evaluator, avoids unexpected collision risks by switching to MPCC when necessary, ensuring the overall safety and reliability of the multi-robot system. The proposed hybrid LSTM method is implemented and tested in the robot operating system (ROS) within diverse constrained scenarios. Experimental results demonstrate that the hybrid LSTM method achieves ≈6% enhancements in trajectory productivity and a reduced computational burden of roughly 75% compared to MPCC, thereby providing a promising solution for local multi-robot trajectory planning in logistics transportation tasks.