The dynamics of synchronization of two-stage van der Pol generator

Results of numerical simulation of self-oscillations synchronization process in two-cascade ring generator van der Pol by harmonic signal are presented. Studies were carried out within the framework of the DT- model of the dynamic system. The model was developed on the basis of the principle of compliance within the framework of the method of slowly changing amplitudes of characteristics of a discrete system with characteristics of an analog prototype. Shortened equations for complex oscillation amplitudes in generator stages are obtained. It was found that in an autonomous system there is an effect of bistability of amplitudes. In the synchronization mode with an external harmonic signal, solutions of shortened equations made it possible to calculate amplitude-frequency and phase-frequency characteristics of synchronous oscillations. It is shown that transitions between bistable states are observed in the synchronous oscillation holding band. Differences of frequency characteristics of synchronization of classical and two-stage oscillators van der Pol were analyzed.

Fast Synchronization Method of Comb-Actuated MEMS Mirror Pair for LiDAR Application

MEMS-based LiDAR (micro-electro–mechanical system based light detection and ranging), with a low cost and small volume, becomes a promising solution for the two-dimensional (2D) and three-dimensional (3D) optical imaging. A semi-coaxial MEMS LiDAR design, based on a synchronous MEMS mirror pair, was proposed in our early study. In this paper, we specifically reveal the synchronization method of the comb-actuated MEMS mirror pair, including the frequency, amplitude, and phase synchronization. The frequency sweeping and phase adjustment are simultaneously implemented to accelerate the MEMS mirror synchronization process. The experiment is set up and the entire synchronization process is completed within 5 s. Eventually, a one-beam MEMS LiDAR system with the synchronous MEMS mirror pair is set up and a LiDAR with a field of view (FOV) of 60°, angular resolution of 0.2°, and frame rate of 360 Hz is obtained. The experimental results verify the feasibility of the MEMS mirror synchronization method and show a promising potential application prospect for the MEMS LiDAR system.

Deploying and scaling distributed parallel deep neural networks on the Tianhe-3 prototype system

Due to the increase in computing power, it is possible to improve the feature extraction and data fitting capabilities of DNN networks by increasing their depth and model complexity. However, the big data and complex models greatly increase the training overhead of DNN, so accelerating their training process becomes a key task. The Tianhe-3 peak speed is designed to target E-class, and the huge computing power provides a potential opportunity for DNN training. We implement and extend LeNet, AlexNet, VGG, and ResNet model training for a single MT-2000+ and FT-2000+ compute nodes, as well as extended multi-node clusters, and propose an improved gradient synchronization process for Dynamic Allreduce communication optimization strategy for the gradient synchronization process base on the ARM architecture features of the Tianhe-3 prototype, providing experimental data and theoretical basis for further enhancing and improving the performance of the Tianhe-3 prototype in large-scale distributed training of neural networks.

Condensation of Synchronous Orbits on Complex Networks

In this paper we study frequency synchronization of Kuramoto oscillators. We find a typical phenomenon of condensed synchronous orbits on single-layer or duplex networks through statistical mechanics analysis and numerical simulations, where the distribution of synchronous orbits is in a bell-shaped form. Further, we investigate phase synchronization on single-layer and duplex networks with different distributions of inherent frequencies. We find that normally distributed inherent frequencies with low variances are more beneficial for phase synchronization, and separately distributed inherent frequencies can slow down the synchronization process. In the end, we investigate the influence of one layer's inherent frequencies on the other layer's phase synchronization through inter-layer couplings. Interestingly, we find that one layer's inherent frequencies with a highly condensed distribution can greatly improve phase synchronization on the other layer. The results shed new lights to our understanding of the nature of synchronization on single-layer as well as multilayer complex networks of coupled Kuramoto oscillators.

Clock synchronization in distributed multicomputer systems. Part II

The study substantiates the necessity of clock synchronization in distributed multicomputer systems. The basic definitions related to the concept of clock synchronization are given, and methods of clock synchronization are classified. Increasing the lifecycle of failure- and fault-tolerant distributed multicomputer systems for critical application is one of the most urgent problems at the current level of technology development. This is especially true for unattended distributed multicomputer systems for space applications. The second part deals with synchronization in systems with Byzantine faults, and this is a complex task due to the characteristic features of the fault model. The synchronization process in multi-cluster and multi-complex systems is associated with the multitasking of such systems, which makes the synchronization process even more relevant and multi-criteria. The paper considers the modern technologies providing the synchronization process in systems of critical use.

Clock synchronization in distributed multicomputer systems. Part I

The study substantiates the necessity of clock synchronization in distributed multicomputer systems. The basic definitions related to the concept of clock synchronization are given, and methods of clock synchronization are classified. Increasing the lifecycle of failure- and fault-tolerant distributed multicomputer systems for critical application is one of the most urgent problems at the current level of technology development. This is especially true for unattended distributed multicomputer systems for space applications. The development of such systems should begin with the construction of models of faults and self-controlled degradation, ensuring, firstly, their failure and fault tolerance and, secondly, maximum survivability, which is possible only if there are means of clock synchronization in such systems. All activities associated with ensuring the synchronization of any distributed multicomputer systems begin with the concept of synchronization of on-board functions, which is based on the generation of on-board time and includes the synchronization of on-board software and equipment that requires time synchronization or information about the course of time. The main elements of this concept are the processor clock module, the onboard software clock, the atomic navigation clock. The first part of the work gives basic definitions, and considers methods and algorithms related to the clock synchronization process. The second part is devoted to synchronization in systems with Byzantine faults and in multi-cluster and multi-complex, i.e. multitask, systems. The modern technologies providing the synchronization process in such systems are considered.