IET Cybersystems and Robotics最新文献

Exploring unknown environments by multiple robots is promising but challenging. The challenges are posed not only by the coordination among multiple robots to improve exploration efficiency, but also by dynamic obstacles that suddenly appear on planned paths. To address those two challenges, this paper proposes a two-layer architecture where the high-level layer generates target locations for each robot to explore the unknown environment, while the low-level layer plans paths in the dynamic environment for each robot. Specifically, in the high-level design, a novel auction algorithm is proposed, which considers both the distance of robots to target locations and the number of frontiers within the clustering domain of target locations. This approach enables robots to explore different target locations while reducing redundant exploration compared to traditional exploration algorithms. In the low-level design, a neural network-based Q-learning algorithm is employed for path planning to achieve dynamic obstacle avoidance. Robots can dynamically adjust their actions through interaction with the external environment, thus avoid obstacles and reach the target position. To validate our methods, a series of simulation experiments are conducted. The experimental results demonstrate that robots can not only efficiently accomplish exploration tasks in unknown environments, but also achieve effective obstacle avoidance when faced with suddenly appearing dynamic obstacles.

Obstructive sleep apnoea (OSA) is a prevalent condition that can lead to various cardiovascular and cerebrovascular diseases, such as coronary heart disease, hypertension, and stroke, posing significant health risks. Polysomnography (PSG) is widely regarded as the most reliable method for detecting sleep apnoea (SA), but it is limited by a complex acquisition process and high costs. To address these issues, some studies have explored the use of single-lead signals, although they often result in lower accuracy due to noise-related information loss. Time context information has been applied to mitigate this issue, but it can lead to overfitting and category confusion. This paper introduces a novel approach utilising time sequence contrastive learning with single-lead electrocardiogram (ECG) signals to detect SA events and assess OSA severity. The proposed method features a Transformer encoder fusion module and a contrastive classification module. First, a multi-branch architecture is employed to extract features from different time scales of the ECG signal, which aids in SA detection. To further enhance the network's focus on the most relevant extracted features, a channel attention mechanism is incorporated to fuse features from different branches. Finally, contrastive learning is utilised to constrain the features, resulting in improved detection performance. A series of experiments were conducted on a public dataset to validate the effectiveness of the proposed method. The method achieved an accuracy of 91.50%, a precision of 92.06%, a sensitivity of 94.37%, a specificity of 86.89%, and an F1 score of 93.20%, outperforming state-of-the-art detection techniques.

Mapless navigation refers to the task of searching for a collision free path without relying on a pre-defined map. Most current works of mapless navigation assume accurate ground-truth localisation is available. However, this is not true, especially for indoor environments, where simultaneous localisation and mapping (SLAM) is needed for location estimation, which highly relies on the richness of environment features. In this work, we propose a novel deep reinforcement learning (DRL) based mapless navigation method without relying on the assumption of the availability of localisation. Our method utilises RGB-D based Oriented FAST and Rotated BRIEF (ORB) SLAM2 for robot localisation. Our policy effectively guides the robot's movement towards the target while enhancing robot pose estimation by considering the quality of the observed features along the selected paths. To facilitate policy training, we propose a compact state representation based on the spatial distributions of map points, which enhances the robot's awareness of areas with reliable map points. Furthermore, we suggest incorporating the relative pose error into the reward function. In this way, the policy will be more responsive to each single action. In addition, rather than utilising a pre-set threshold, we adopt a dynamic threshold to improve the policy's adaptability to variations in SLAM performance across different environments. The experiments in localisation challenging environments have demonstrated the remarkable performance of our proposed method. It outperforms the related DRL based methods in terms of success rate.

This paper presents a visual simultaneous localization and mapping (SLAM) system designed for highly dynamic environments, capable of eliminating dynamic objects using only visual information. The proposed system integrates learning-based and geometry-based methods to address the challenges posed by moving objects. The learning-based approach leverages image segmentation to remove previously trained objects, whereas the geometry-based approach utilises point correlation to eliminate unseen objects. By complementing each other, these methods enhance the robustness of the SLAM system in dynamic scenarios. Experimental results demonstrate that the proposed method effectively removes dynamic objects. Comparative studies with state-of-the-art algorithms further show that the proposed method achieves superior accuracy and robustness.

Given the complexity of live working environments in power distribution networks, where autonomous obstacle avoidance by robots often involves numerous path nodes and low exploration efficiency, the Bidirectional Node-Controlled Rapidly Exploring Random Tree (BNC-RRT) algorithm is proposed. This algorithm guides path search by progressively altering the sampling area and employs a node control mechanism to constrain the random tree expansion and extract effective boundary points. This approach reduces the number of ineffective nodes and collision checks during the search process, thereby enhancing path planning efficiency. Comparative simulation experiments conducted in various scenarios demonstrate that this algorithm reduces the number of path nodes and improves planning efficiency compared to classical algorithms. Finally, real-world experiments on a live working robot developed by our team show that the proposed algorithm shortens the average path length by 8.6%, and reduces the average planning and movement times by 44.7% and 28.7%, respectively, compared to classical path planning algorithms. These results indicate that the algorithm effectively improves path planning efficiency and is suitable for live working tasks in the power distribution industry.

This paper introduces a pioneering dynamic system optimisation for multiagent (DySOMA) framework, revolutionising task scheduling in dynamic intelligent spaces with an emphasis on multirobot systems. The core of DySOMA is an advanced auction-based algorithm coupled with a novel task preemption ranking mechanism, seamlessly integrated with an ontology knowledge graph that dynamically updates. This integration not only enhances the efficiency of task allocation among robots but also significantly improves the adaptability of the system to environmental changes. Compared to other advanced algorithms, the DySOMA algorithm shows significant performance improvements, with its RLB 26.8% higher than that of the best-performing Consensus-Based Parallel Auction and Execution (CBPAE) algorithm at 10 robots and 29.7% higher at 20 robots, demonstrating its superior capability in balancing task loads and optimising task completion times in larger, more complex environments. DySOMA sets a new benchmark for intelligent robot task scheduling, promising significant advancements in the autonomy and flexibility of robotic systems in complex evolving environments.

With the booming development of logistics, manufacturing and warehousing fields, the autonomous navigation and intelligent obstacle avoidance technology of automated guided vehicles (AGVs) has become the focus of scientific research. In this paper, an enhanced deep reinforcement learning (DRL) framework is proposed, aiming to empower AGVs with the ability of autonomous navigation and obstacle avoidance in the unknown and variable complex environment. To address the problems of time-consuming training and limited generalisation ability of traditional DRL, we refine the twin delayed deep deterministic policy gradient algorithm by integrating adaptive noise attenuation and dynamic delayed updating, optimising both training efficiency and model robustness. In order to further strengthen the AGV's ability to perceive and respond to changes of a dynamic environment, we introduce a distance-based obstacle penalty term in the designed composite reward function, which ensures that the AGV is capable of predicting and avoiding obstacles effectively in dynamic scenarios. Experiments indicate that the AGV model trained by this algorithm presents excellent autonomous navigation capability in both static and dynamic environments, with a high task completion rate, stable and reliable operation, which fully proves the high efficiency and robustness of this method and its practical value.

Spike transformers cannot be pretrained due to objective factors such as lack of datasets and memory constraints, which results in a significant performance gap compared to pretrained artificial neural networks (ANNs), thereby hindering their practical applicability. To address this issue, we propose a hybrid attention spike transformer that utilises self-attention with compound tokens and channel attention-based token processing to better capture the inductive biases of the data. We also add convolution in patch splitting and feedforward networks, which not only provides local information but also leverages the translation invariance and locality of convolutions to help the model converge. Experiments on static datasets and neuromorphic datasets demonstrate that our method achieves state-of-the-art performance in the spiking neural networks (SNNs) field. Notably, we achieve a top-1 accuracy of 80.59% on CIFAR-100 with only 4 time steps. As far as we know, it is the first exploration of the spike transformer with multiattention fusion, achieving outstanding effectiveness.

Accurate target detection in low-light environments is crucial for unmanned aerial vehicles (UAVs) and autonomous driving applications. In this study, the authors introduce a real-time multimodal fusion for enhanced detection (RMF-ED), a novel framework designed to overcome the limitations of low-light target detection. By leveraging the complementary capabilities of near-infrared (NIR) cameras and light detection and ranging (LiDAR) sensors, RMF-ED enhances detection performance. An advanced NIR generative adversarial network (NIR-GAN) model was developed to address the lack of annotated NIR datasets, integrating structural similarity index measure (SSIM) loss and L1 loss functions. This approach enables the generation of high-quality NIR images from RGB datasets, bridging a critical gap in training data. Furthermore, the multimodal fusion algorithm integrates RGB images, NIR images, and LiDAR point clouds, ensuring consistency and accuracy in proposal fusion. Experimental results on the KITTI dataset demonstrate that RMF-ED achieves performance comparable to or exceeding state-of-the-art fusion algorithms, with a computational time of only 21 ms. These features make RMF-ED an efficient and versatile solution for real-time applications in low-light environments.

A swarm of unmanned aerial vehicles (UAVs) has been widely used in both military and civilian fields due to its advantages of high cost-effectiveness, high task efficiency and strong survivability. However, there are still challenges in flocking control of UAV swarms in complex environments with various obstacles. In this paper, we propose a flocking control and obstacle avoidance method for UAV swarms, which is called willingness control method (WCM). Specifically, we propose an adaptive risk avoidance willingness (ARAW) mechanism, in which each UAV has an ARAW coefficient representing its ARAW. As the distance from danger gets closer, the ARAW of the UAV to avoid danger increases. On this basis, an obstacle avoidance method for UAV swarms is designed, and an informed individual mechanism influenced by neighbour repulsion is introduced. By combining the hierarchical weighting Vicsek model (HWVEM), the UAV swarm system can simultaneously balance flocking navigation and obstacle avoidance tasks and adjust the priority of different tasks adaptively during the task process. Finally, under local communication constraints of the UAV, a series of simulation experiments as well as real-word experiments with up to 12 UAVs are conducted to verify the security and compactness of the proposed method.