Search

In this article, we propose a solution to multi-satellite intelligent task planning using the multi-agent reinforcement learning (MARL) method. Fristly, we have developed a multi-satellite task planning model based on the Markov game framework. Furthermore, we have computationally designed a satellite state transition function to address the task planning problem and successfully solved it using the multi-agent proximal policy optimization (MAPPO) algorithm. Our experimental results demonstrate that the MARL method exhibits remarkable convergence speed and performance, delivering significant rewards in multi-scale task planning scenarios. Consequently, it proves to be a highly suitable approach for multi-satellite intelligent task planning.

Target detection technology has been greatly improved for synthetic aperture radar (SAR) images recently, due to the advancement in the deep learning domain. However, because of the existence of clutter in the SAR images, it is still a challenge to detect small targets with high accuracy and low computational complexity. To solve this problem, a detection algorithm based on a feature fusion and cross-layer connection network is proposed in this article. First, attention feature fusion is applied to improve the feature fusion ability for the small targets by allocating weights to various feature maps adaptively. Meanwhile, the depthwise separable convolution (DW-Conv) is used to reduce the computational complexity caused by the increasement of network layers. Then, a cross-layer connection (Cross-Connect) submodule is proposed to fuse shallow features with deep features further. Finally, a multiscale target detection (Multi-Detect) submodule is designed to improve the detection ability for small targets. We compare the proposed algorithm with the other representative methods on the SAR-Ship-Dataset and SSDD, quantitative evaluations show that our proposed algorithm can reach the highest computational efficiency. Therefore, because of the superior performance in terms of accuracy and efficiency, the algorithm proposed in this article is more suitable to detect small targets for the SAR images.

In order to meet the rapid growth of railway transportation demand, the technology that can improve its capacity has been widely concerned. Virtual Coupling (VC) is a new technology to improve capacity by decreasing headways between successive trains. Its basic principle is to control the train formation operation in coordination with the goal of keeping the same speed under the support of data transmission technology such as train-to-train communication. Our paper first reviews the existing theoretical research on train formation operation control, which is divided into four categories: train following, feedback control, optimal control and computational intelligent method. Secondly, by reviewing the related projects in Europe and China, based on the scenario analysis method, five general scenarios and two emergency scenarios in the whole process of VC operation are sorted out. Then, an original, complete list of performance indicators (PIs) is proposed for evaluating the performance of VC in different scenarios. Additionally, the paper gives a brief insight into VC operation by providing an adequate context to understand the proposed PIs.

In recent years, virtual synchronous generator (VSG) control algorithm has been widely studied in distributed grid-connected power inverters to deal with active and reactive power adjustment by simulating the operation principle of the actual synchronous generator. Nevertheless, once the current sensor fails, the grid-connected power inverter will lose its stability. To enhance the fault-tolerant operation ability of the grid-connected power inverter, a new inverter side current sensor-less model predictive control method based on sliding mode observer (SMO) is proposed. The stability of the SMO is analyzed by using the Lyapunov stability criterion. The gains of the SMO are then designed by analyzing their influences on the current and voltage observation based on the closed-loop transfer functions. Besides, to show the influences of the capacitance changes on the current observation, a detailed parameter sensitivity analysis is also carried out. Then, by using the estimated inductance current from the SMO, a current sensor-less model predictive control method is finally proposed. The main novelty of the proposed method is that both the output voltage of the inverter and the inverter side inductance are not required in the proposed SMO. So, the accuracy of current sensor-less control is improved. To verify the effectiveness of the proposed method, detailed experimental studies are carried out based on a Typhoon and PE-Expert 4 based experimental platform.

In attosecond and strong-field physics, the acquisition of few-cycle laser sources open up a new area. We report a nonlinear pulse compression technology combining multi-thin-plate spectral broadening and nonlinear self-compression that generates a 0.52 mJ, good spatial quality characteristics and a spectral bandwidth supporting a 14 fs Fourier transform limited duration at 1 kHz repetition rate and at a center wavelength of 1.9 μm. The total energy transfer efficiency is up to 83%. Pulse to pulse stability of the energy output is 0.7% (RMS). The pulse duration is near 3 optical cycles. This pulse compression approach can be a key-enabling technology for the next generation of extreme photonics, attosecond research and coherent x ray-science, and it also can be further extended to mid-infrared lasers with longer wavelengths and higher peak power.

The width of convolutional neural networks (CNNs) is crucial for improving performance. Many wide CNNs use a convolutional layer to fuse multiscale features or fuse the preceding features to subsequent features. However, these CNNs rarely use blocks, which consist of a series of successive convolutional layers, to fuse multiscale features. In this paper, we propose an approach for improving performance by fusing the low-level features extracted from different blocks. We utilize five different convolutions, including $3\times 3, 5\times 5, 7\times 7,5\times 3\,\cup \, 3\times 5$ and $7\times 3\,\cup \, 3\times 7$ , to generate five low-level features, and we design two fusion strategies: low-level feature fusion (L-Fusion) and high-level feature fusion (H-Fusion). Experimental results show that the L-Fusion is more helpful for improving the performance of CNNs, and the $5\times 5$ convolution is more suitable for multiscale feature fusion. We summarize the conclusion as a strategy that fuses multiscale features in the preceding stage of CNNs. Furthermore, we propose a new architecture to perceive the input of CNNs by using two self-governed blocks based on the strategy. Finally, we modify five off-the-shelf networks, DenseNet-BC (depth = 40), ALL-CNN-C (depth = 9), Darknet 19 (depth = 19), Resnet 18 (depth = 18) and Resnet 50 (depth = 50), by utilizing the proposed architecture to verify the conclusion, and these updated networks provide more competitive results.

Constrained multi-objective optimization problems are common in practical engineering and are more difficult to handle than unconstrained problems. In general, it is necessary to find a balance between the convergence and diversity of solutions, as well as its feasibility. For the constrained multi/many-objective optimization problem, a particle swarm optimization algorithm based on a two-level balance strategy is proposed. In contrast to existing views, the first level of the proposed algorithmic framework emphasizes convergence, while diversity and feasibility are considered together as the second-level scheme. An ensemble fitness ranking was used to improve the convergence of the proposed algorithm. To balance the diversity and solution feasibility, the solutions are selected by combining the angles between the solutions using the constraint dominance principle. A penalty-based boundary-crossing approach is used as a utility function to calculate the fitness of the populations, which is compared with six state-of-the-art constrained multi/many-objective evolutionary optimization algorithms on multiple constrained test suites, and the experimental results show that the proposed algorithm is highly competitive in most test problems. Furthermore, to illustrate the effect of different utility functions on the performance of the algorithm, the Chebyshev decomposition method is employed and compared with the former, and the results show that different utility functions need to be chosen to cope with problems of different characteristics.

Recently, deep convolutional neural networks (CNNs) have been widely explored in single image super-resolution(SISR) and obtained remarkable performance. However, most of the existing CNN-based SISR methods tend to produce over-smoothed outputs and miss some textural details. To address these issues, we propose a wavelet-based asymmetric convolution network (WACN). Different from conventional CNN methods that directly infer HR images, our approach firstly learns to predict the LR’s corresponding series of HR’s wavelet coefficients before reconstructing HR images from them. This helps to capture more structural information in images to preserve texture information and avoid artifacts. To enhance the ability of feature extraction, we propose an asymmetric convolution block (ACB) structure to form a very deep network. In the training phase, ACB can provide different receptive fields to enrich feature information. In the inference phase, ACB’s asymmetric convolution kernel can be equivalently fused into the standard square-kernel layers, such that no extra computational burdens are introduced in the inference phase. Furthermore, we propose a variance-based channel attention (VCA) mechanism to adaptively rescale channel-wise features by considering interdependencies among channels. Extensive experimental results demonstrate the superiority of the proposed WACN in comparison with the state-of-the-art methods.

A near-infrared (NIR) supercontinuum and ultrashort pulses are generated via phase-mismatched cascaded frequency conversion in DSTMS (4-N, N-dimethylamino-4'-N'-methyl-stilbazolium 2, 4, 6-trimethylbenzenesulfonate) crystal. NIR pulses with a central wavelength of 1.5 μm and 63 fs pulse duration are incident into the crystal, and the spectrum is broadened to over 1200-2100 nm. Compensating the dispersion by fused silica, ultrashort NIR pulses with 16.6 fs pulse duration are obtained. Meanwhile, intense single-cycle terahertz pulses are generated via optical rectification in DSTMS crystal.

This paper focuses on the inverse synthetic aperture radar (ISAR) imaging of nonuniformly rotating targets under low signal-to-noise ratio (SNR) environment, and an effective ISAR imaging algorithm based on a novel parameter estimation method is proposed. For nonuniformly rotating targets, the received signals in a range cell can be modeled as multicomponent cubic phase signals (CPSs) after the migration compensation and translation-induced phase error correction. For the CPS model, the chirp rate (CR) and quadratic chirp rate (QCR) will induce image defocus. To estimate the CR and QCR of CPS, a novel parameter method is presented. The basic idea of this method is to design a third order autocorrelation function (TOAF) which enable coherent accumulation of the signal energy in both of the time and lag-time domain. Compared with several state-of-the-art algorithms, the proposed method has a significant anti-noise performance improvement, while offering a proper computational complexity. The effectiveness and the superiority of the proposed ISAR imaging algorithm have been demonstrated through several simulation results.

Over the past decades, plenty of filtering algorithms have been presented to distinguish ground and non-ground points from airborne LiDAR point clouds. However, with the existing methods, it is difficult to derive satisfactory filtering results on rugged terrains with dense vegetation due to the low-level penetration ability of laser pulses. Therefore, a multi-level interpolation-based filter is developed in this paper. The novelty of the algorithm lies within its usage of multi-scale morphological operations and robust z-score to correctly select ground seeds as more as possible, and a terrain-adaptive residual threshold to adapt to various terrain characteristics. Rural samples provided by International Society for Photogrammetry and Remote Sensing (ISPRS) were employed to assess the performance of the proposed method and its results were compared with 15 filtering algorithms developed in recent 5 years (2015-2019). Results show that the proposed method with the optimized parameters produces the best accuracy with the average total error and kappa coefficient of 1.89% and 87.88%, respectively. We further filtered high-density point clouds in six forested areas with different vegetation covers and terrain slopes. Results demonstrate that the proposed algorithm is more accurate than the well-known filtering methods including morphological-based filter, progressive TIN densification filter (PTD), improved PTD and cloth simulation filter, with the average total error decreased by 26.2%, 19.9%, 3.8% and 40.4%, respectively. Moreover, the DEMs of the proposed method have lower average root mean square errors than the four classical filters. Therefore, the proposed method can be considered as an effective ground filtering algorithm for airborne LiDAR point clouds in forested areas.

The advanced driving assistant system (ADAS) is an important vehicle safety technology that can effectively reduce traffic accidents. This system can perceive information about the surrounding environment through in-vehicle cameras. However, these cameras are easily affected by severe weather conditions, such as those involving fog, rain, and snow. The quality of the images acquired by the system is degraded, and the function of the ADAS is thus weakened. In response to this problem, we propose a comprehensive imaging model that can represent the features of fog, rain streaks, raindrops and snowflakes in an image. Subsequently, an algorithm called RASWNet is proposed, which can remove all severe weather features from a degraded image. Based on the generative adversarial network, RASWNet combines the focus capture ability of a visual attention mechanism, the memory ability of the recurrent neural network and the feature extraction ability of the dense blocks approach. We verify the network structure through several ablation studies and use various synthetic and real images to test it. The results of these experiments show that our algorithm is not only better than the commonly used algorithms in terms of its clarity enhancement capacity but is also suitable for all severe weather conditions.

The oxygen content of boiler flue gas is a valid indicator of boiler efficiency and emissions. Measuring the oxygen content of boiler flue gas is time consuming and costly. To overcome the latter shortcomings, a novel deep belief network algorithm based hybrid prediction model for the oxygen content of boiler flue gas is proposed. First, the algorithm is used to build a model based on the historical data collected from the distribution control system. The variables are divided into control variables and state variables to meet the needs of advanced control requirement. Then, a lasso algorithm is used to select variables highly related to the oxygen content as the inputs of the prediction model. Two basic models based on the deep-belief network are established, one using control variables, and the other, state variables. Finally, the two basic models are combined with a least square support vector machine to improve prediction accuracy of the oxygen content of boiler flue gas. To test the accuracy of the proposed algorithm, experiments based on three industrial datasets are performed. Performance of the comparison of the proposed deep belief algorithm is compared with five machine learning algorithms. Computational experience has shown that the model derived with the deep-belief algorithm produced better accuracy than the models generated by the other algorithms.

In this paper, a more generalized dilated nested array (DNA) named multi-level dilated nested array (multi-level DNA), which further extends the original DNA to the general situation and improves the direction of arrival (DOA) estimation performance, is proposed. The original DNA, which consists of a nested array and a time-shifted array, possesses the capabilities of handling more sources and achieving improved estimation accuracy compared with the nested array. However, only one time-shifted array is employed in DNA and the performance improvement is quite limited. Here, a new multi-level DNA, which utilizes multiple time-shifted nested arrays, is proposed. The multi-level DNA can provide more degrees of freedom (DoFs), meanwhile its difference co-array is still hole-free. The closed form expression of uniform DoFs in multi-level DNA is derived. The corresponding DOA Cramér-Rao bound, which gives the low bound on the variance of estimated DOA, is deduced in detail and some properties related to it are concluded. In total, for the DOA estimation, multi-level DNA can detect more sources and achieve more accurate estimation performance compared with the original DNA. Numerical simulations are performed to verify the superiority of the proposed multi-level DNA.

Line feature constitutes important geometric structure information in image processing and is significant in visual navigation and 3-D structure contour extraction of 3-D objects. The current state-of-the-art line extraction methods exhibit fast extracting speed and good extracting results. However, most line extraction methods face a problem, wherein a long line is easily fragmented. Thus, a long line is broken into several short segments due to local small changes or disturbances. To obtain longer and more useful line segments without the aid of external information, this paper proposes an improved line extraction method called affine-lines, which is based on an affine camera model. First, the affine camera model is used to simulate affine projections from different viewing angles, and a sequence of affine simulated images is obtained via affine transformation matrices. Subsequently, line segments are extracted from the original image and each simulated image. Each set of line segments on the simulated image is back-projected on the original image based on its corresponding inversely affine transformation matrix. Finally, the line segments on the original image are used as references to sequentially purify and optimize subsequent lines. Several defined geometric constraints are used to eliminate pseudo lines and combine short lines. Thus, fragmented lines are connected, and short lines are converted into long lines. Given several sets of close-range and aerial images, experiments are performed and compared via a state-of-the-art line extraction algorithm, called the line segment detector (LSD). The results indicate that the proposed method significantly increases the lengths of the obtained line segments, significantly reduces the fragmentation effect, and obtains more useful line segments. The extracted improved line segments are applied to line matching for building structures, and promising results are obtained.

The high noise levels of high-rate Global Navigation Satellite System (GNSS) solutions limit their seismological applications, including capturing earthquake-induced coseismic displacements. In this study, we developed a new adaptive denoising approach for high-rate GNSS observations to improve the precision of seismic displacements and preserve seismic waveforms in GNSS coseismic signals. The performance of the proposed method was evaluated using high-rate (5 Hz) GNSS data acquired from the moderate EI Mayor-Cucapah earthquake (Mw 7.2) on 4 April 2010 and the small Brawley seismic swarm (Mw 4.6-5.5) on 26 August 2012. The performance of the proposed method was compared with those of modified sidereal filtering, Stacking filtering, and MSF plus Stacking. The comparison showed that the proposed method is more precise than other widely used method. It can significantly remove high-frequency white noise and low-frequency colored noise caused by CME, multipath errors, and/or other unmodeled systematic errors in high-rate GNSS displacements. The results were also compared with collocated strong motion data (50 and 200 Hz). The high precision of the proposed method was mainly afforded by the high performance of complete ensemble empirical mode decomposition, which was used to decompose the GNSS signal into different frequency modes. However, the normalized autocorrelation function and correlation coefficients used to determine the noise-dominated high-frequency modes and the “wavelet-like” soft-threshold used for direct denoising of the noise-dominated high-frequency modes also contributed. Despite the high noise levels of GNSS solutions, especially regarding the vertical displacement components, some small-amplitude details, which are usually only detectable by seismic instruments, could be observed in the denoised displacements in this study. The results reported herein indicate that the proposed method significantly improves the precision and reliability of GNSS displacements and the effectiveness of seismic signal detection, which is particularly critical for the measurement of earthquake-induced coseismic displacements.

A model predictive control-based heuristic dynamic programming (HDP-MPC) online strategy is proposed in this paper, to further improve the fuel economy for the series hybrid loaders. The back propagation neural network (BPNN) is utilized to establish the dynamics model for series hybrid loader, to effectively characterize its real working process. Next, a HDP-MPC approximate optimal algorithm is proposed. And its convergence and the controlled system stability are separately proved. Besides, an online energy management strategy based on HDP-MPC is designed to implement the energy management controller. Finally, the BPNN dynamics model and online HDP-MPC strategy are simulated and validated. Through the experimental results analysis, contrasted with the analytic model, the established dynamics model can more effectively reflect the actual motion process of the series hybrid loader with higher accuracy. Moreover, compared with the fuzzy rule-based and adaptive equivalent consumption minimization strategy strategies, the designed HDP-MPC strategy can respectively achieve fuel-savings of 28.65% and 10.38%, supplying a new idea for online energy management of the series hybrid loaders.