IET Cybersystems and Robotics最新文献

The problem of triangular lattice formation in robot swarms has been investigated extensively in the literature, but the existing algorithms can hardly keep comparative performance from swarm simulation to real multi-robot scenarios, due to the limited computation power or the restricted field of view (FOV) of robot sensors. Eventually, a distributed solution for triangular lattice formation in robot swarms with minimal sensing and computation is proposed and developed in this study. Each robot is equipped with a sensor with a limited FOV providing only a ternary digit of information about its neighbouring environment. At each time step, the motion command is directly determined by using only the ternary sensing result. The circular motions with a certain level of randomness lead the robot swarms to stable triangular lattice formation with high quality and robustness. Extensive numerical simulations and multi-robot experiments are conducted. The results have demonstrated and validated the efficiency of the proposed approach. The minimised sensing and computation requirements pave the way for massive deployment at a low cost and implementation within swarms of miniature robots.

The task of few-shot object detection is to classify and locate objects through a few annotated samples. Although many studies have tried to solve this problem, the results are still not satisfactory. Recent studies have found that the class margin significantly impacts the classification and representation of the targets to be detected. Most methods use the loss function to balance the class margin, but the results show that the loss-based methods only have a tiny improvement on the few-shot object detection problem. In this study, the authors propose a class encoding method based on the transformer to balance the class margin, which can make the model pay more attention to the essential information of the features, thus increasing the recognition ability of the sample. Besides, the authors propose a multi-target decoding method to aggregate RoI vectors generated from multi-target images with multiple support vectors, which can significantly improve the detection ability of the detector for multi-target images. Experiments on Pascal visual object classes (VOC) and Microsoft Common Objects in Context datasets show that our proposed Few-Shot Object Detection via Class Encoding and Multi-Target Decoding significantly improves upon baseline detectors (average accuracy improvement is up to 10.8% on VOC and 2.1% on COCO), achieving competitive performance. In general, we propose a new way to regulate the class margin between support set vectors and a way of feature aggregation for images containing multiple objects and achieve remarkable results. Our method is implemented on mmfewshot, and the code will be available later.

In this study, a robust model predictive controller is designed for the trajectory tracking problem of non-holonomic constrained wheeled mobile robot based on an elliptic invariant set approach. The controller is based on a time-varying error model of robot kinematics and uses linear matrix inequalities to solve the robust tracking problem taking uncertainties into account. The uncertainties are modelled by linear fractional transform form to contain both parameter perturbations and external disturbances. The control strategy consists of a feedforward term that drives the centre of the ellipse to the reference point and a feedback term that converges the uncertain system state error to the equilibrium point. The strategy stabilises the nominal system and ensures that all states of the uncertain system remain within the ellipsoid at each step, thus achieving robust stability of the uncertain system. Finally, the robustness of the algorithm and its resistance to disturbances are verified by simulation and experiment.

In recent years, deep learning has been applied to a variety of scenarios in Industrial Internet of Things (IIoT), including enhancing the security of IIoT. However, the existing deep learning methods utilised in IIoT security are manually designed by heavily relying on the experience of the designers. The authors have made the first contribution concerning the joint optimisation of neural architecture search and hyper-parameters optimisation for securing IIoT. A novel automated deep learning method called synchronous optimisation of parameters and architectures by GA with CNN blocks (SOPA-GA-CNN) is proposed to synchronously optimise the hyperparameters and block-based architectures in convolutional neural networks (CNNs) by genetic algorithms (GA) for the intrusion detection issue of IIoT. An efficient hybrid encoding strategy and the corresponding GA-based evolutionary operations are designed to characterise and evolve both the hyperparameters, including batch size, learning rate, weight optimiser and weight regularisation, and the architectures, such as the block-based network topology and the parameters of each CNN block. The experimental results on five intrusion detection datasets in IIoT, including secure water treatment, water distribution, Gas Pipeline, Botnet in Internet of Things and Power System Attack Dataset, have demonstrated the superiority of the proposed SOPA-GA-CNN to the state-of-the-art manually designed models and neuron-evolutionary methods in terms of accuracy, precision, recall, F1-score, and the number of parameters of the deep learning models.

The multiple travelling salesman problem (mTSP) is a classical optimisation problem that is widely applied in various fields. Although the mTSP was solved using both classical algorithms and artificial neural networks, reiteration is inevitable for these methods when presented with new samples. To meet the online and high-speed logistics requirements deploying new information technology, the iterative algorithm may not be reliable and timely. In this study, a deep convolutional neural network (DCNN)-based solution method for mTSP is proposed, which can establish the mapping between the parameters and the optimal solutions directly and avoid the use of iterations. To facilitate the DCNN in establishing a mapping, an image representation that can transfer the mTSP from an optimisation problem into a computer vision problem is presented. While maintaining the excellent quality of the results, the efficiency of the solution achieved by the proposed method is much higher than that of the traditional optimisation method after training. Meanwhile, the method can be applied to solve the mTSP under different constraints after transfer learning.

A novel distributed control architecture for unmanned aircraft system (UASs) based on the new Robot Operating System (ROS) 2 middleware is proposed, endowed with industrial-grade tools that establish a novel standard for high-reliability distributed systems. The architecture has been developed for an autonomous quadcopter to design an inclusive solution ranging from low-level sensor management and soft real-time operating system setup and tuning to perception, exploration, and navigation modules orchestrated by a finite-state machine. The architecture proposed in this study builds on ROS 2 with its scalability and soft real-time communication functionalities, while including security and safety features, optimised implementations of localisation algorithms, and integrating an innovative and flexible path planner for UASs. Finally, experimental results have been collected during tests carried out both in the laboratory and in a realistic environment, showing the effectiveness of the proposed architecture in terms of reliability, scalability, and flexibility.

In this study, the state estimation problems for linear discrete systems with uncertain parameters, deterministic input signals and d-step measurement delay are investigated. A robust state estimator with a similar iterative form and comparable computational complexity to the Kalman filter is derived based on the state augmentation method and the sensitivity penalisation of the innovation process. It is discussed that the steady-state properties such as boundedness and convergence of the robust state estimator under the assumptions that the system parameters are time invariant. Numerical simulation results show that compared with the Kalman filter, the obtained state estimator is more robust to modelling errors and has nice estimation accuracy.

Accurate and robust positioning and mapping are the core functions of autonomous mobile robots, and the ability to analyse and understand scenes is also an important criterion for the intelligence of autonomous mobile robots. In the outdoor environment, most robots rely on GPS positioning. When the signal is weak, the positioning error will interfere with the mapping results, making the semantic map construction less robust. This research mainly designs a semantic map construction system that does not rely on GPS signals for large outdoor scenes. It mainly designs a feature extraction scheme based on the sampling characteristics of Livox-AVIA solid-state LiDAR. The factor graph optimisation model of frame pose and inertial measurement unit (IMU) pre-integrated pose, using a sliding window to fuse solid-state LiDAR and IMU data, fuse laser inertial odometry and camera target detection results, refer to the closest point distance and curvature for semantic information. The point cloud is used for semantic segmentation to realise the construction of a 3D semantic map in outdoor scenes. The experiment verifies that laser inertial navigation odometry based on factor map optimisation has better positioning accuracy and lower overall cumulative error at turning, and the 3D semantic map obtained on this basis performs well.

In this study, a localisation system without cumulative errors is proposed. First, depth odometry is achieved only by using the depth information from the depth camera. Then the point cloud cross-source map registration is realised by 3D particle filtering to obtain the pose of the point cloud relative to the map. Furthermore, we fuse the odometry results with the point cloud to map registration results, so the system can operate effectively even if the map is incomplete. The effectiveness of the system for long-term localisation, localisation in the incomplete map, and localisation in low light through multiple experiments on the self-recorded dataset is demonstrated. Compared with other methods, the results are better than theirs and achieve high indoor localisation accuracy.

This study addresses two issues about the interaction of the upper limb rehabilitation robot with individuals who have disabilities. The first step is to estimate the human's target position (also known as TPH). The second step is to develop a robust adaptive impedance control mechanism. A novel Non-singular Terminal Sliding Mode Control combined with an adaptive super-twisting controller is being developed to achieve this goal. This combination's purpose is to provide high reliability, continuous performance tracking of the system's trajectories. The proposed adaptive control strategy reduces matched dynamic uncertainty while also lowering chattering, which is the sliding mode's most glaring issue. The proposed TPH is coupled with adaptive impedance control with the use of a Radial Basis Function Neural Network, which allows a robotic exoskeleton to simply track the desired impedance model. To validate the approach in real-time, an exoskeleton robot was deployed in controlled experimental circumstances. A comparison study has been set up to show how the adaptive impedance approach proposed is better than other traditional controllers.