Search

Performance of supercomputers has been steadily and exponentially increasing for the past 20 years, and is expected to increase further. This unprecedented computational power enables us to build and simulate large-scale neural network models composed of tens of billions of neurons and tens of trillions of synapses with detailed anatomical connections and realistic physiological parameters. Such “human-scale” brain simulation could be considered a milestone in computational neuroscience and even in general neuroscience. Towards this milestone, it is mandatory to introduce modern high-performance computing technology into neuroscience research. In this article, we provide an introductory landscape about large-scale brain simulation on supercomputers from the viewpoints of computational neuroscience and modern high-performance computing technology for specialists in experimental as well as computational neurosciences. This introduction to modeling and simulation methods is followed by a review of various representative large-scale simulation studies conducted to date. Then, we direct our attention to the cerebellum, with a review of more simulation studies specific to that region. Furthermore, we present recent simulation results of a human-scale cerebellar network model composed of 86 billion neurons on the Japanese flagship supercomputer K (now retired). Finally, we discuss the necessity and importance of human-scale brain simulation, and suggest future directions of such large-scale brain simulation research.

Computer simulation of the human brain at an individual neuron resolution is an ultimate goal of computational neuroscience. The Japanese flagship supercomputer, K, provides unprecedented computational capability toward this goal. The cerebellum contains 80% of the neurons in the whole brain. Therefore, computer simulation of the human-scale cerebellum will be a challenge for modern supercomputers. In this study, we built a human-scale spiking network model of the cerebellum, composed of 68 billion spiking neurons, on the K computer. As a benchmark, we performed a computer simulation of a cerebellum-dependent eye movement task known as the optokinetic response. We succeeded in reproducing plausible neuronal activity patterns that are observed experimentally in animals. The model was built on dedicated neural network simulation software called MONET (Millefeuille-like Organization NEural neTwork), which calculates layered sheet types of neural networks with parallelization by tile partitioning. To examine the scalability of the MONET simulator, we repeatedly performed simulations while changing the number of compute nodes from 1,024 to 82,944 and measured the computational time. We observed a good weak-scaling property for our cerebellar network model. Using all 82,944 nodes, we succeeded in simulating a human-scale cerebellum for the first time, although the simulation was 578 times slower than the wall clock time. These results suggest that the K computer is already capable of creating a simulation of a human-scale cerebellar model with the aid of the MONET simulator.

As friction couples tangential and lateral degrees-of-freedom of a structure at contact interfaces, the resulting asymmetric dynamic system is prone to dynamic instability. Using state-feedback control, such a frictional asymmetric system can be stabilized through assigning the system desirable eigenvalues; but uncertainties in system parameters can cause assigned eigenvalues to deviate from desired locations and thus stability may be lost. This study presents a robust stabilization method that assigns both desirable eigenvalues and their sensitivities and thus render assigned eigenvalues stable and insensitive to perturbations in uncertain contact parameters (the friction coefficient, contact damping, and contact stiffness). This method utilizes receptances of the corresponding symmetric part of the asymmetric system. The optimal control input location is first determined by minimizing the Frobenius norm of the normalized eigen-sensitivity matrix. The normalized eigen-sensitivities indicate that the friction coefficient and contact stiffness intrinsically have similar crucial effects on the stability of the system. To demonstrate the application of the proposed control method, the eigen-sensitivities with respect to only the friction coefficient are assigned. A constrained over-determined least-squares problem is solved to assign both required eigenvalues and eigen-sensitivities. Numerical examples validate the effectiveness of the proposed robust control scheme by Monte Carlo simulations.

One of the grand challenges for computational neuroscience and high-performance computing is computer simulation of a human-scale whole brain model with spiking neurons and synaptic plasticity using supercomputers. To achieve such a simulation, the target network model must be partitioned onto a number of computational nodes, and the sub-network models are executed in parallel while communicating spike information across different nodes. However, it remains unclear how the target network model should be partitioned for efficient computing on next generation of supercomputers. Specifically, reducing the communication of spike information across compute nodes is essential, because of the relatively slower network performance than processor and memory. From the viewpoint of biological features, the cerebral cortex and cerebellum contain 99% of neurons and synapses and form layered sheet structures. Therefore, an efficient method to split the network should exploit the layered sheet structures. In this study, we indicate that a tile partitioning method leads to efficient communication. To demonstrate it, a simulation software called MONET (Millefeuille-like Organization NEural neTwork simulator) that partitions a network model as described above was developed. The MONET simulator was implemented on the Japanese flagship supercomputer K, which is composed of 82,944 computational nodes. We examined a performance of calculation, communication and memory consumption in the tile partitioning method for a cortical model with realistic anatomical and physiological parameters. The result showed that the tile partitioning method drastically reduced communication data amount by replacing network communication with DRAM access and sharing the communication data with neighboring neurons. We confirmed the scalability and efficiency of the tile partitioning method on up to 63,504 compute nodes of the K computer for the cortical model. In the companion paper by Yamaura et al., the performance for a cerebellar model was examined. These results suggest that the tile partitioning method will have advantage for a human-scale whole-brain simulation on exascale computers.

An optimal state feedback controller based on H2 norm is proposed for attenuating the belt vibration caused by loading disturbance in a class of image transfer belt which is widely applied in multi-color printers. Based on the simplified mechanical model of the image transfer belt unit, the presented controller acts on the actuating torque generated by a direct-current motor and optimizes the defined variables related to the quality of the original image formation. Simulation results responding to a periodic loading disturbance ascertain its efficacy with comparisons of the conventional speed control. It is certified that the image transfer belt unit is successfully maintained in a satisfying condition with effective vibration attenuation under the proposed control architecture.

An optimal control architecture based on $$H_\infty $$H? norm is proposed for rejection of loading disturbance in a class of image transfer belt system, modeled as a four-rollers-belt powered by a direct-current motor which is widely applied in printers and copiers. The presented controller acts on the actuating torque from the motor and optimizes all variables concerning quality of the original image formation with human visual transfer effect. Simulation results responding to a periodic loading disturbance with comparison of the insufficiency and performance limitation in conventional speed control ascertain its effectiveness. It is shown that the image transfer belt system is successfully maintained in the satisfying condition with significant vibration reductions.

In recent years computer has become a tool that cannot be separated from our life. As long as we are working on computers, the comfort of the workstation become important since improper position leads to repetitive strain injuries. In this research avoiding discomfort is approached by making a workstation capable of changing working positions. So, a new workstation concept was designed to create a comfortable computer work experience in which a user can work in different working position at a single workstation. In this paper the design of the footrest and its effect in comfort of workstation is discussed. The footrest had 3 degrees of freedom and its positions were controlled by three linear actuators. Evaluation results in four different working positions showed that flexible footrest has a great impact on the comfort of seated computer work. It increased comfort in thigh, leg and foot relative to a standard workstation without footrest.

Dopaminergic systems have been known to be involved in the regulation of locomotor activity and development of psychosis. However, the observations that some Parkinson's disease patients can move effectively under appropriate conditions despite low dopamine levels (eg, kinesia paradoxia) and that several psychotic symptoms are typical antipsychotic resistant and atypical antipsychotic sensitive indicate that other systems beyond the dopaminergic system may also affect locomotor activity and psychosis. The present study showed that dopamine-deficient (DD) mice, which had received daily L-DOPA injections, could move effectively and even be hyperactive 72 h after the last L-DOPA injection when dopamine was almost completely depleted. Such hyperactivity was ameliorated by clozapine but not haloperidol or ziprasidone. Among multiple actions of clozapine, muscarinic acetylcholine (ACh) activation markedly reduced locomotor activity in DD mice. Furthermore, the expression of choline acetyltransferase, an ACh synthase, was reduced and extracellular ACh levels were significantly reduced in DD mice. These results suggest that the cholinergic system, in addition to the dopaminergic system, may be involved in motor control, including hyperactivity and psychosis. The present findings provide additional evidence that the cholinergic system may be targeted for the treatment of Parkinson's disease and psychosis.

This paper investigates the dynamics of the giant swing motions of an underactuated three-link gymnastic robot moving in a vertical plane by means of dynamic delayed feedback control (DDFC). DDFC, being one of useful methods to overcome the so-called odd number limitation in controlling a chaotic discrete-time system, is extended to control a continuous-time system such as a 3-link gymnastic robot with passive joint. Meanwhile, a way to calculate the error transfer matrix and the input matrix which are necessary for discretization is proposed, based on a Poincare section which is defined to regard the target system as a discrete-time system. Moreover, the stability of the closed-loop system by the proposed control strategy is discussed. Furthermore, some numerical simulations are presented to show the effectiveness in controlling a chaotic motion of the 3-link gymnastic robot to a periodic giant swing motion.

During voluntary limb movements, humans exert anticipatory postural adjustments (APAs) to prevent any upcoming equilibrium disturbance that might be provoked by limb movements. Dysfunction in generation or control of APAs is associated with postural deficits in some human patients with cerebellar damage. To examine the role of the cerebellum in APAs, we investigated a conditional transgenic mouse of spinocerebellar ataxia type 3 (SCA3Tg) that has defective cerebellar Purkinje cells. Kinematic analyses and monitoring of electromyographic activities during quadrupedal standing showed that SCA3Tg mice exhibited greater hindlimb instability than wild-type (WT) mice. This instability increased during a reaching task that required postural adjustments associated with voluntary neck movements. Normally, the activities of the hindlimb muscles are synchronized with those in the neck that are the agonists for movement of the head in this reaching task; however, in SCA3Tg mice, activities in the hindlimbs were markedly delayed compared to the neck. These observations cannot simply be explained as a secondary outcome of the muscle atrophy that occurs in SCA3Tg mice. In WT mice with muscle atrophy induced by immobilization of the hindlimbs, we did not find impairment of APAs. These findings suggest that the deficits in APAs during the reaching task in SCA3Tg mice were not due to muscle atrophy in the hindlimbs, but were mainly caused by cerebellar degeneration. Therefore, we conclude that the cerebellum is critically involved in APAs.

This paper presents the parameter optimization procedure for a new anthropomorphic finger mechanism in performing pinching and self-adaptive grasping operation. The mechanism is made based on the observation of human finger during grasping and pinching. It is constructed from seven bar linkages and its upper middle phalanx is incorporated with a curved guiding slot and lead screw mechanism to adjust the link's effective length in accomplishing these two tasks. The finger's geometrical parameters are optimized based on four design criteria, which are the anthropomorphic pinching kinematics and force; and grasping kinematics and force to enhance its performance. The numerical result obtained gives an improved parameter set compared to the initial guess. A prototype with small size and light weight actuators and sensors has been implemented based on the optimum solution to achieve a compact anthropomorphic design with grasping and pinching capabilities.

This paper discusses a control method which can stabilize the giant swing motions of 3-link horizontal bar gymnastic robot. A method called Multiple-prediction Delayed Feedback Control(MDFC) proposed by the authors, which has been shown to be effective in control such chaos system, is extended in this paper to experimental gymnastic robot system, taking in account the effect of friction around the link joint. The dynamic of underactuated systems like gymnastic robot shows significant differences in cases of with friction and without friction. However, MDFC considers only the ideal situation without friction. Therefore, a stabilization control method consisting of three kind of control inputs which the one is MDFC for guaranteeing asymptotic stability and the others are used for friction compensation is derived. Numerical simulations and experimental results prove its effectiveness of our proposed method.

Currently, equipments using paper feeding mechanism such as copy-machine and printer are required higher paper feeding velocity. It is necessary to reveal the impact of paper feeding velocity on the characteristics of paper feeding mechanism. In this paper, high speed paper feeding experiment was firstly performed. And the characteristics of paper feeding mechanism dependences on paper feeding velocity were verified. Next, the effects of rubber viscoelasticity, dynamic friction coefficient of rubber dependences on slipping velocity and dynamic friction coefficient of paper dependences on slipping velocity were shown with experiments and two dimensional analyses. It is found that the characteristics of paper feeding mechanism dependences on paper feeding velocity are mainly caused by dynamic friction coefficient of paper dependences on slipping velocity.

This paper studies the behavior of the giant swing motions of a 3-link horizontal bar gymnastic robot using the delayed feedback control (DFC). A modified DFC method called Multiple-prediction Delayed Feedback Control (MDFC), which controls the chaotic system via periodic gain is proposed in this paper. First, plural Poincare maps are defined to regard the target continuous-time system as a T-periodic discrete-time system, so that the system stability can be evaluated based on the theory of monodromy matrix. Second, a way to calculate analytically the error transfer matrix and the input matrix which are necessary for discretization is presented. Finally, simulation results show that our proposed method is effective in the control of the giant swing motions.

This paper presents a new Feedback Error Learning (FEL) scheme with the application of Radial Basis Function Network (RBFN) based Finite Impulse Response (FIR) filter to control underactuated systems. This method provides a stable feedback controller which is systematically derived from sequential backstepping control procedure. Besides, it also formulates a simple approach for FEL feedforward controller structure by employing the inverse dynamic model of the plant with physical parameters. The RBFN based FIR filter is used in obtaining the estimation of the state variables to produce an ideal feedforward control input. Simulation results on a two link acrobat robot with nonzero initial angular momentum in achieving a final desired posture angle are presented to show the validity of the proposed algorithm.

System control and identification often require accurate estimations of velocity and acceleration with a reduced group delay and low-pass property to attenuate high frequency noise. This paper proposes a novel design method of FIR filter for estimating velocity and acceleration from sampled displacement data, where the filter coefficients are derived by piecewise Radial Basis Function Network (RBFN). The proposed FIR filter not only provides more accurate estimates when compared to the conventional method, but also achieves the prescribed low-pass property and specified group delay in the pass band. The theoretical analysis of RBFN in the frequency domain reveals the relationship between the frequency response of the FIR filter and width/regularization parameters in the RBFN. Finally, the procedures for designing the filter are summarized with design examples.

This paper discusses the seek control for head positioning mechanism with a contact slider in hard disk drives (HDDs).In General, to increase the control force is the most popular step used to decrease the seek time; it often results however in transient vibration of HDDs that has start and stop motion. For high-speed, high-precision seeking mechanisms, it is necessary to apply control methods which induce no residual vibration at a target track even with a large control force. The proposed positioning control system consists of two-degrees-of-freedom controller. The optimal point-to-point (PTP) control with a frequency-shaped weighting function is employed in feedfoward input design to eliminate vibratory responses due to mechanical vibration modes. To estimate and compensate the friction force of the contact slider generated by the contact of a disk and a slider, a equivalent disturbance estimate observer is designed. The validity of the proposed approach has been verified by long span seek simulation trials in the benchmark problems for HDDs.

This paper describes a new Dynamic Absorber Suspension (DAS) structure which has a significant damping effect for the flow-induced sway mode and second torsion mode. First we designed the DAS structure for a practically used suspension by Finite Element Analysis. From frequency response function of DAS, we found that magnification factor of sway mode can be reduced to 14 dB, and that of second torsion mode is 5 dB. Next, we fabricated the DAS and measured the frequency characteristics by laser Doppler vibrometer using a hydrostatic air-bearing jig and a 3.5 inch disk spindle. As a result, we found that the DAS has a remarkable damping effect on the sway mode and second torsion mode compared with a conventional suspension with a damper sheet.

Friction induced vibration (FIV) attracts extensive attention of researchers due to its ubiquitousness and intriguing complexity. Disc brake squeal is deemed one of the typical FIV problems with eigenvalues of the brake system locating on the right-hand side half of the complex plane. In this research, an active state feedback control strategy is proposed for shifting the unstable eigenvalues of the brake system to the stable region. The presented control method is constructed based on the receptance method in which the knowledge of mass, damping and stiffness matrices is not required. Moreover, the robustness of the control method is explored against uncertainty in friction coefficient. The numerical simulation results demonstrate that the proposed method successfully places the required poles of the system and thus stabilises an unstable system in the presence of an uncertain parameter.Copyright © 2015 by ASME