Ising machines are physical simulators implemented in various unconventional hardware that aim at reducing the energy consumption cost and time needed to study the properties of complex disordered systems. Many complex non-convex (or rugged) energy landscapes, notably the Ising Hamiltonian of binary spin systems, can be mapped directly onto these physical simulators. We focus on analog Ising machines, which state is described by real variables, and discuss reduction in time and energy-consumption of newly proposed dynamics. The developed simulator will constitute a new opportunity for exploring the physics of complex disordered systems.
Silicon neuronal networks have been gathering increasing attention as a promising platform for the next generation of artificial intelligence, because they can realize neuronal information processing with low power consumption, on account of their parallel and distributed structures. We have studied about the Digital Spiking Silicon Neuron model that is designed to reproduce a wide variety of neuronal activities with less necessary circuit resources. In this paper, we briefly introduce the DSSN model, especially about its parameter tuning method.
Persistent homology computes the change of the homology groups of a shape growing as a parameter increases. Persistent homology groups are defined for filtrations, which are increasing sequence of simplicial complexes. Vietoris-Rips complexes are filtrations constructed from point clouds. Persistent homology groups are computed with a variation of a reduction algorithm in linear algebra. An example of the computation is given in this article.
The Boolean satisfiability (SAT) problem, which is a problem of finding a satisfying assignment of given Boolean formulae, is an important problem both in theory and practice. In this article, the dynamical system solving the SAT problem proposed by Ercsey-Ravasz and Toroczkai is described. This system reduces the SAT problem as the time-varying optimization problem and solves it using the gradient descent method in continuous time and state space. It has interesting dynamical properties, such as transient chaos, and also has possible application as electric circuit implementation.
Machine learning has attracted much attention in science and engineering. In particular, Bayes estimation is one of the most important methods in machine learning and variational Bayes （VB） inference is a widely used algorithm for it. On the other hand, computations based on quantum mechanics also have attracted much attention since they are expected to lead to a breakthrough. In this review paper, we explain a recently proposed quantum-mechanical extension of VB. We also discuss two kinds of problem settings: unsupervised and supervised learning.
Mathematical models of epidemic diseases are used for understanding complex spreading processes of diseases and assessing the efficacy of countermeasures to epidemics. Among them, metapopulation epidemic models consider populations distributed in spatially distant patches and migration of individuals between patches. These models can be theoretically analyzed to some extent as well as can incorporate realistic factors. In this review, we focus on local interventions to a fraction of patches for reducing the effective transmission rates in the patches. We explain the analytical derivation of the minimum fraction of patches (or the intervention threshold） that is necessary for prevention of global epidemic outbreaks and numerical simulations for validating the efficacy of targeted intervention.
We describe the features of Fourier transform and recurrence plot, that are both reversible transformations for time series. Then we discuss the possibility of the recurrence plot as an operator for time series. The idea is to define a basis of time series including nonlinearity, and classify the points in the given time series through a distance of the points.
Timeline statistics of the accumulated total counts of the deceased and infected with COVID-19 in China from January through March 2020 has revealed a characteristic feature in the transitional lethal rate （accumulated total count of the dead divided by the accumulated total count of confirmed infection） that has a temporally local minimum sometime eight day before the appearance of the peak of the fatality. Such a characteristic is also found in the theoretical model that uses fitting sigmoid functions.
Infection control policies are expected to suppress the cumulative number of cases. In this research, we analyzed effects of infection control policies on the final size using the SIR model. We derived the final size equation in the presence of the policy intervention, and showed that the final size decreases compared with that without the intervention. We found that the final size may not necessarily be minimized when the policy intervention starts at the initial stage of the infection process. A numerical result supporting this fact was presented.
Nowadays, a lot of people expect the reduction of the cost of the infrastructure, including traffic signals. We discuss and propose the control logic of autonomous traffic signal system, which is one of the answers of the problem regarding the cost. In the proposed logic, the best signal pattern is selected from three points of view, such as safety, smoothness and equality. We can apply the logic to every signalized intersection regardless of the intersection shape and the signal type. The signals based on the logic work sufficiently, but are independent of the central control system.
Freeway-on ramp merging sections cause interactions between merging and through vehicles: each vehicle’s driver decides on his action based on the traffic conditions and other drivers’ behavior. This paper models twovehicle interaction on the merging section as a non-cooperative game. We assume that two drivers are players in the game under complete information, where each player tries to maximize his utility function. The utility functions depend on distance and speed difference between two vehicles. To estimate the parameters of the utility functions, we used Zen Traffic Data collected in Hanshin Expressway and applied the recently proposed statistical method.
It is highly expected that by deployment of automated driving systems, some social problems are solved or improved such as supporting the people who cannot reach the transportation system. The objective of this paper is to clarify how automated driving vehicles to be deployed for solving or improving such social problems. In this paper, barriers in social deployment of automated driving vehicles would be identified by using issue tree, considering many stakeholders (e.g., car manufacturers, public sectors, service users and insurance companies). Also, strategies to remove or improve these barriers would be discussed.
This study conducts an empirical analysis of the capacity drop phenomenon at sag bottlenecks based on the recent advanced continuum traffic flow theory. We first show the underlying phenomena, components, and consequences of the theory in a concise manner. We then compare several theoretical predictions with real data, qualitatively and quantitatively, which demonstrates that the theory brings us new interpretations of observations.
Origin-destination (OD) traffic volume represents an important parameter of traffic demand, when considering more efficient traffic management strategies on expressway network. In this study, we analyzed temporal fluctuation of OD traffic volume in the whole Tokyo metropolitan expressway network, using the ETC data which is recorded by electronic toll collection system at tollgates. To be more specific, periodical component by time in a day was extracted from the observed data for each OD by applying state space method, and which was categorized by time series clustering. As a result, we found ten typical patterns of temporal fluctuation with different peak periods.
Since 2015, when the author published open public transportation data in the GTFS format in Shizuoka Prefecture, the number of open data publishers is rapidly increasing. Standardization of data format by the Ministry of Land, Infrastructure, Transport and Tourism in 2017 is supporting this trend. The role of the data is becoming more critical in Recent MaaS (Mobility as a Service) movement. In this paper, I will introduce the latest state of open public transportation data and discuss technical issues and future possibilities.
Platoon system is expected as a solution to problems such as fuel efficiency, lack of driver and improvement of safety. To implement the system, it is necessary to ensure safety and acceptability of peripheral drivers. In this research, we aim to improve the safety and social acceptability of platoon system. We examined the effect of specification of truck platooning and infrastructure such as distance between the platooning trucks, external HMI of platooning trucks, the alert sign on the road and length of acceleration lane and so on, on the driving behaviors and the acceptability of peripheral drivers using driving simulator.
In a remote-controlled vehicle system, a pilot operates in an environment where we cannot feel the physical sensation of being inside the vehicle, and the system has a time lag. Therefore, remote vehicle control must not be considered as an extension of car driving; in contrast, it should be considered as a new type of mobility. This study aims to clarify the driving characteristics of a remote vehicle pilot and to develop the human machine interface (HMI) for the remote-control system. In this paper, we conducted a basic study of HMI for a remote-control system using a driving simulator.
We developed a method to recognize such hand signals from on-vehicle camera, based on deep-learning technique. The skeleton coordinate of the performer is input to a deep learning method, to classify the signal state into Red/Green or Red/Green/Other. From the state and the continuation conditions, the instruction Stop/Go is determined. Our preliminary experiment proved that quite similar short actions are included both in Red and Green, and it is better to separate such actions as “Other”. In the final result, Stop/Go can be appropriately determined, and at the same time, the temporal difference of estimation between switching Stop/Go （Too-early Go and Too-late Stop） was less than 0.43 seconds.
Traffic accidents has seriously jeopardized the safety of human. Among all causes, poor visual environment in blind spots like intersections is a critical reason. Trying to solve such problem, we have been developing a method to detect approaching objects in road safety mirror using on-vehicle camera. We use an on-vehicle camera to input driving scenes in front of vehicle, then road safety mirror itself and the road objects reflected in it are detected from driving scenes using Faster R-CNN. Subsequently, each road object is tracked by using detection results, with Kalman filter to interpolate interrupted tracks and optical flow to determine whether to end interpolation. The vertical moving direction of objects in road safety mirror is used to judge whether an object is coming near to or going away from observer. We evaluate our method using self-made road safety mirror and road objects datasets, which presents that while larger road objects are detected with high precision and recall, tiny objects are still hard to detect and track.
The railway crossing control system that sounds alarm and closes crossing gates by sending the signal to the controller from the center server calculating the fastest time of the railway vehicle to reach the crossing has been produced. The self-localization of the vehicle is done receiving GNSS signals and obtained information is transmitted to the center server using LTE cell phone line. Taking advantage of versatility of LTE line, the unified control system for railway crossings and traffic signals on road are proposed. The performance of the proposed system is shown through the experiments carried out at the experimental field inside the campus of the university.
Disposing the significant amounts of concrete wastes and CO2 emissions produced from cement industries is a major challenge in fields pertaining to concrete as well as society. To address this issue, wooden waste was utilized in this research to recycle concrete waste without cement. Concrete and wooden wastes were crushed, milled, mixed, and hot pressed to produce concrete recycled with wood (CRW). Parameters such as the mixing proportion of concrete waste and wood, moisture content, and temperature of hot pressing were varied to study the effect of these parameters on the bending strength of the CRW. The results showed that a majority of the CRWs prepared under various conditions exhibited higher bending strengths as compared to conventional concrete.
Curing agents are often used to improve the finishability of concrete surfaces. This study aims to quantitatively evaluate the influence of a surfactant-based curing agent on the viscoelasticity of cement paste via the free damped oscillation method, using a rigid-body pendulum. The results recorded indicate that the relative storage modulus and the relative loss modulus of cement paste are computable from the oscillation cycle and logarithmic decrement of the free damped oscillation of a rigid-body pendulum. Mixing the surfactant-based curing agent into cement paste yielded a more significant effect than mixing water, delaying and restraining the increase in the relative storage modulus and relative loss modulus of the cement paste.