Frontiers in Neurorobotics最新文献

The transportation of hazardous chemicals on roadways has raised significant safety concerns. Incidents involving these substances often lead to severe and devastating consequences. Consequently, there is a pressing need for real-time detection systems tailored for hazardous material vehicles. However, existing detection methods face challenges in accurately identifying smaller targets and achieving high precision. This paper introduces a novel solution, HMV-YOLO, an enhancement of the YOLOv7-tiny model designed to address these challenges. Within this model, two innovative modules, CBSG and G-ELAN, are introduced. The CBSG module's mathematical model incorporates components such as Convolution (Conv2d), Batch Normalization (BN), SiLU activation, and Global Response Normalization (GRN) to mitigate feature collapse issues and enhance neuron activity. The G-ELAN module, building upon CBSG, further advances feature fusion. Experimental results showcase the superior performance of the enhanced model compared to the original one across various evaluation metrics. This advancement shows great promise for practical applications, particularly in the context of real-time monitoring systems for hazardous material vehicles.

Insects exhibit remarkable abilities in navigating complex natural environments, whether it be evading predators, capturing prey, or seeking out con-specifics, all of which rely on their compact yet reliable neural systems. We explore the field of bio-inspired robotic vision systems, focusing on the locust inspired Lobula Giant Movement Detector (LGMD) models. The existing LGMD models are thoroughly evaluated, identifying their common meta-properties that are essential for their functionality. This article reveals a common framework, characterized by layered structures and computational strategies, which is crucial for enhancing the capability of bio-inspired models for diverse applications. The result of this analysis is the Strategic Prototype, which embodies the identified meta-properties. It represents a modular and more flexible method for developing more responsive and adaptable robotic visual systems. The perspective highlights the potential of the Strategic Prototype: LGMD-Universally Prototype (LGMD-UP), the key to re-framing LGMD models and advancing our understanding and implementation of bio-inspired visual systems in robotics. It might open up more flexible and adaptable avenues for research and practical applications.

With the development of deep learning, synthetic aperture radar (SAR) ship detection and recognition based on deep learning have gained widespread application and advancement. However, there are still challenging issues, manifesting in two primary facets: firstly, the imaging mechanism of SAR results in significant noise interference, making it difficult to separate background noise from ship target features in complex backgrounds such as ports and urban areas; secondly, the heterogeneous scales of ship target features result in the susceptibility of smaller targets to information loss, rendering them elusive to detection. In this article, we propose a context-aware one-stage ship detection network that exhibits heightened sensitivity to scale variations and robust resistance to noise interference. Then we introduce a Local feature refinement module (LFRM), which utilizes multiple receptive fields of different sizes to extract local multi-scale information, followed by a two-branch channel-wise attention approach to obtain local cross-channel interactions. To minimize the effect of a complex background on the target, we design the global context aggregation module (GCAM) to enhance the feature representation of the target and suppress the interference of noise by acquiring long-range dependencies. Finally, we validate the effectiveness of our method on three publicly available SAR ship detection datasets, SAR-Ship-Dataset, high-resolution SAR images dataset (HRSID), and SAR ship detection dataset (SSDD). The experimental results show that our method is more competitive, with AP50s of 96.3, 93.3, and 96.2% on the three publicly available datasets, respectively.

Insulators play a pivotal role in the reliability of power distribution networks, necessitating precise defect detection. However, compared with aerial insulator images of transmission network, insulator images of power distribution network contain more complex backgrounds and subtle insulator defects, it leads to high false detection rates and omission rates in current mainstream detection algorithms. In response, this study presents ID-YOLOv7, a tailored convolutional neural network. First, we design a novel Edge Detailed Shape Data Augmentation (EDSDA) method to enhance the model's sensitivity to insulator's edge shapes. Meanwhile, a Cross-Channel and Spatial Multi-Scale Attention (CCSMA) module is proposed, which can interactively model across different channels and spatial domains, to augment the network's attention to high-level insulator defect features. Second, we design a Re-BiC module to fuse multi-scale contextual features and reconstruct the Neck component, alleviating the issue of critical feature loss during inter-feature layer interaction in traditional FPN structures. Finally, we utilize the MPDIoU function to calculate the model's localization loss, effectively reducing redundant computational costs. We perform comprehensive experiments using the Su22kV_broken and PASCAL VOC 2007 datasets to validate our algorithm's effectiveness. On the Su22kV_broken dataset, our approach attains an 85.7% mAP on a single NVIDIA RTX 2080ti graphics card, marking a 7.2% increase over the original YOLOv7. On the PASCAL VOC 2007 dataset, we achieve an impressive 90.3% mAP at a processing speed of 53 FPS, showing a 2.9% improvement compared to the original YOLOv7.

Loop closure detection is an important module for simultaneous localization and mapping (SLAM). Correct detection of loops can reduce the cumulative drift in positioning. Because traditional detection methods rely on handicraft features, false positive detections can occur when the environment changes, resulting in incorrect estimates and an inability to obtain accurate maps. In this research paper, a loop closure detection method based on a variational autoencoder (VAE) is proposed. It is intended to be used as a feature extractor to extract image features through neural networks to replace the handicraft features used in traditional methods. This method extracts a low-dimensional vector as the representation of the image. At the same time, the attention mechanism is added to the network and constraints are added to improve the loss function for better image representation. In the back-end feature matching process, geometric checking is used to filter out the wrong matching for the false positive problem. Finally, through numerical experiments, the proposed method is demonstrated to have a better precision-recall curve than the traditional method of the bag-of-words model and other deep learning methods and is highly robust to environmental changes. In addition, experiments on datasets from three different scenarios also demonstrate that the method can be applied in real-world scenarios and that it has a good performance.

Target assignment and path planning are crucial for the cooperativity of multiple unmanned aerial vehicles (UAV) systems. However, it is a challenge considering the dynamics of environments and the partial observability of UAVs. In this article, the problem of multi-UAV target assignment and path planning is formulated as a partially observable Markov decision process (POMDP), and a novel deep reinforcement learning (DRL)-based algorithm is proposed to address it. Specifically, a target assignment network is introduced into the twin-delayed deep deterministic policy gradient (TD3) algorithm to solve the target assignment problem and path planning problem simultaneously. The target assignment network executes target assignment for each step of UAVs, while the TD3 guides UAVs to plan paths for this step based on the assignment result and provides training labels for the optimization of the target assignment network. Experimental results demonstrate that the proposed approach can ensure an optimal complete target allocation and achieve a collision-free path for each UAV in three-dimensional (3D) dynamic multiple-obstacle environments, and present a superior performance in target completion and a better adaptability to complex environments compared with existing methods.