Search

The parallel computing capabilities of GPUs have a significant impact on computationally intensive iterative tasks. Offloading part or all of the deep learning tasks from the CPU to the GPU for execution is mainstream. However, a large number of redundant iterative calculations exist in the iterative process of computing tasks. Therefore, we propose a GPU-based distributed incremental iterative computing architecture that can make full use of distributed parallel computing and GPU memory structure. The architecture supports deep learning and other computationally intensive iterative applications by optimizing data placement and reducing redundant iterative calculations. To support block-based data partitioning and coalesced memory access on GPUs, we propose GDataSet, an abstract data set. The GPU incremental iteration manager called GTracker is designed to be responsible for GDataSet cache management on the GPU. In order to solve the limitation of on-chip memory size, we propose a variable sliding window mechanism. It improves the hit rate of cache access and the speed of data access by realizing the best block arrangement between on-chip memory and off-chip memory. Besides, a communication channel based on an incremental iterative model is designed to support data transmission and task communication in cluster computing. Finally, we implement the proposed architecture based on Spark 2.4.1 and CUDA 10.0. Comparative experiments with widely used computationally intensive iterative applications (K-means, LSTM, etc.) show that the incremental iterative acceleration architecture can significantly improve the efficiency of iterative computing.

Scheduling plays an important role in improving the performance of big data-parallel processing. Spark is an in-memory parallel computing framework that uses a multi-threaded model in task scheduling. Most Spark task scheduling processes do not take the memory into account, but the number of concurrent task threads determined by the user. It emerges as a potential limitation for the performance. To overcome the limitations in the Spark-core source code, this paper proposes a dynamic Spark memory-aware task scheduler (DMATS), which not only treats memory and network I/O as a computational resource but also dynamically adjusts concurrency when scheduling tasks. Specifically, we first analyze the RDD based Spark execution engine to obtain the amount of task processing data and propose an algorithm for estimating the initial adaptive task concurrency, which is integrated with the known task input information and the executor memory. Then, a dynamic adjustment algorithm is proposed to change the concurrency dynamically through feedback information to optimally utilize the limited memory resources. We implement a dynamic memory-aware task scheduling (DMATS) in Spark 2.3.4 and evaluate performance with two typical benchmarks, shuffle-light and shuffle-heavy. The results show that the algorithm not only reduces the execution time by 43.64%, but also significantly improves resource utilization. Experiments also show that our proposed method has advantages compared with other similar works such as WASP.

Stepper motor control system includes driver, motor, rotating mechanism and mechanical load, which requires high precision control. Predictive maintenance of stepper motor control system is conducive to improving the stability of the system. In this paper, predictive maintenance analysis is carried out for the common problems of mechanical load increase and shaft misalignment. The predictive maintenance platform of stepper motor is established. Through predictive maintenance means, the running status of stepping motor is monitored, and the load value signal and position residual signal parameters of stepping motor are collected and analyzed to realize real-time perception of health status and trend prediction, and timely judge the problem of shaft misalignment and load increase of stepping motor, so as to formulate reasonable predictive maintenance The strategy can improve the stability and reliability of the whole system.

In order to more visually display and analyze the temporal and spatial evolution of coronavirus disease (COVID- 19), and provide refined data support and information reference of epidemic prevention decisions for government departments, this paper designs and develops a COVID-19 epidemic simulation and visualization system based on WebGIS through a third-party open-source framework and visualization library. The system can visually portray the microscopic behavior of individuals spreading the epidemic and the macroscopic epidemic situation. The system implements three modules: epidemic visualization, intra-urban travel intensity and individual model transmission simulation. In this paper, the changes of travel intensity within the city and the temporal and spatial evolution of the epidemic were analyzed with Wuhan city as an example. This study is valuable for the development of systems for visualizing outbreaks of highly infectious epidemics.

In this paper, the least square support vector machine(LSSVM) fault diagnosis method based on load value analysis is adopted, the load feedback value is filtered by wavelet packet decomposition, and the eigenvector is extracted. Particle swarm optimization optimizes the kernel parameters c and g of LSSVM algorithm, and the optimal LSSVM model based on load value analysis is obtained. The running fault of stepping motor is predicted by this model. The experimental results show that the optimal LSSVM model based on load value analysis has higher classification accuracy and smaller computation time. It is an effective way to classify step motor faults.

The interest in flapping-wing Micro Air Vehicles (MAVs) has been rising progressively in the past years, as they can combine high agility manoeuvres with precision hovering flight and can be applied in complex spaces for reconnaissance missions. In this study, we propose a double flapping-wing MAV, which has four wings comprised by two pairs, each pair is driven by one brush motor and one linear servo. The flapping mechanism is composed of a crank-rocker and double rocker mechanism, which can amplify the output angle of wings and used for lift increasing. We take the Rhinoceros beetle as a bionic object and the Weis-Fogh mechanism as the high lift generation principle. The vehicle can actively control 4 degrees of freedom (DOFs), namely, roll, pitch, yaw, and thrust. Compare to the single pair counterpart, our vehicle possess a high thrust-to-weight ratio, which make it possible for more onboard load and beneficial to attitude control, additionally, the 3 rotational DOFs (roll, pitch, and yaw) is completely uncoupled and controlled independently, which is useful for control system design. The currently vehicle weighting 32.8 g (without the battery) and can generate 0.34N thrust at the maximum flapping frequency of approximately 23 Hz.

In view of the shortcomings of the previous mechanism design for the flapping wing micro air vehicle(FMAV),such as small flapping angle and low transmission efficiency, a new type mechanism of FMAV is designed in this paper. Large aerodynamic force is generated by the large amplitude flapping angle and high frequency flapping. By combining the new control mechanism, the maneuverability of the aircraft and the load demand during the mission are improved. Meanwhile, this new type mechanism meets the design requirements of the miniaturization and flexibility, and provide the theoretical basis and application value for the mission of the aircraft to carry out the border patrol, the disaster rescue and the reconnaissance missions.

For the rapid growth computation requirements in big data and artificial intelligence area, CPU-GPU heterogeneous clusters can provide more powerful computing capacity compared to CPU clusters. The number of GPUs on single computing node is scalable, which greatly improves the computing capacity of the cluster under the condition of limited cluster size. However, there is a lack of the effective load-balancing scheduling model in multi-GPU hardware environment. This paper proposes AEML, an acceleration engine for multi-GPU load-balancing in distributed heterogeneous environment. AEML can effectively integrate GPUs into distributed processing framework and achieve great load-balance among multiple heterogeneous GPUs. We propose a heterogeneous task execution model based on multiple GPUs and multiple streams (MGMS), which can effectively balance the workload of multiple GPUs. MGMS model utilizes four core techniques: a fine-grained task mapping mechanism, a device resource unified management scheme, a novel resource-aware GPU task scheduling strategy and a feedback-based streams adjustment scheme. The implementation of AEML system is based on Spark 2.4.1 and NVIDIA CUDA 10.0. We comprehensively evaluate the performance of AEML with multiple typical benchmarks. Experimental results show that AEML can fully exploit the computing power of GPUs and achieve great load-balance among multiple heterogeneous GPUs.

Compressive sensing theory can recover the original signal from little observation data, by means of the non-adaptive measurement far below Nyquist sampling and the optimization method. Due to the number of track is limited and the uneven distribution of tracks, the traditional method is difficult to image effectively in the practical application, while the compressive sensing theory by tomoSAR three-dimensional imaging (MUSIC, APES, etc.) can solve the problem easily, and it can realize high resolution imaging of SAR in the height direction. This paper applied the Generalized Orthogonal Matching Pursuit (GOMP) algorithm to the SAR tomography imaging, which can accurately reconstruct the signal in the height direction. Meanwhile, in this paper, 14 images of ascend track from which is “TerraSAR-X” HH polarization mode in Beijing during the period of 2011–2014 were taken as the test data, and Beijing pangu7 star hotel is taken as the research target. Finally, the result indicates that our algorithm is better than typical Orthogonal Matching Pursuit (OMP) algorithm.

Urban building outline is not only a very important land cover type, but also the key of studying urban building. It is significance on urban construction planning and natural disasters monitoring. This paper discussed the method for building information extracting according to high-resolution SAR coherence image. Through theoretical analysis and test verifying, it had obtained a better effect for extracting the buildings’ profiles or top figures from the coherence image, which often expressed as ‘L’-style top structures from SAR SLC image. This method has been proved the further potential in extracting urban buildings’ structure and profile information, and has some extent of reference means.

The teaching presupposition in the simple horizon is mechanical reflex to teaching regularity. The presupposition become “track”, the teaching become the process along the “track”. The Teaching presupposition based on complex theory, examine and reconstruction realizes the horizon conversion from simple to complex.