本記事に「抄録」はありません。

抄録全体を表示

Social simulation attracts many interests as an engineering approach of design and analysis of social systems. Especially, post-peta computing power will enable total design for safety and robustness of social services by exhaustive simulation. This article provides several examples of exhaustive social simulation, and overviews future works to enhance usage of such a technique.

抄録全体を表示

We evaluate the effectiveness of real planning and control about the traffic in Hiroshima city by simulating the traffic flows in the entire city. Our massively parallel simulation of minute behavior of individual drivers is made credible by the use of novel techniques for learning the parameters of simulation models from available data.

抄録全体を表示

Researches on smart mobility are carried out to solve urban problems such as environmental issues and traffic congestion. This paper illustrates some key technologies for modeling and optimizing people flow to realize a smart mobility system focusing on agent simulation. This paper also explains some studies on agent simulation and future prospects.

抄録全体を表示

Multi-Agent Simulation (MAS) could be an efficient approach to consider social measures. However, in order to make the measures more practical and efficient, the simulation model must reflect regional characteristics in detail and its validity should be shown. In this paper, we introduce our work using MAS in the two fields: disaster prevention and transportation demand management (TDM), and show how our model is convincing and practical. For the example of disaster prevention, we introduce Tsunami evacuation simulation in Kamakura-City. In this work, we examined a hearing survey and a field survey to take account of the behavioral characteristic in that area in our simulation model. For the example of TDM, we develop a traffic behavior model considering the decision making process of drivers in a particular region. These factors make our simulation model more persuasive and realistic, and the simulation result is available to plan specific social measures. Moreover, MAS will be useful for consensus forming with residents and users for such social measures since the simulation result is quantitative and can be easily visualized.

抄録全体を表示

In this paper, we evaluate how to provide or control traffic information based on an anticipatory stigmergy. Managing traffic congestion is one of the main issues for smart cities, and many works have been trying to address it from the IT and transportation research perspectives. Recently, dynamic coordination methods are becoming possible using the more short-term traffic information that can be gathered by probe-vehicles or smart phones. Some approaches have been trying to handle short-term traffic information in which a stigmergy-based approach is employed as an indirect communication method for cooperation among distributed agents and for managing traffic congestion. One drawback of these approaches is that handling near-future congestion remains problematic because stigmergies are basically past information. Therefore we propose anticipatory stigmergy for sharing information on near-future location and allocating drivers adequately. All vehicles submit their near-future intention as anticipatory stigmergy to search their routes. Our preliminary results demonstrate that anticipatory stigmergy with assignment strategy works well.

抄録全体を表示

In this paper, we propose an autonomous Micro Grid operation by using multi-agent approach. The proposed multi-agent system consists of seven types of agent: single Micro Grid controller agent (MGC), several Load agents (LAGs), several gas-turbines (gas-engine) agents (GAGs), several photovoltaic generation agents (PVAGs), several wind-turbine generation agents (WTAGs), several battery agents (BAGs), and single grid agent (GridAG). In a Micro Grid, LAGs act as consumers or buyers, GAGs, PVAGs and WTAGs act as producers or sellers, and BAGs act as prosumers or sellers/buyers. In order to verify the performance of the proposed system, it applied to a simple model system with different electrical power prices. From the simulation results, it can be seen the proposed multi-agent system could perform the Smart Grid operation efficiently.

抄録全体を表示

We describe agent-based simulations to accomplish the levelization of excessive electric power loads which arise from the connection of electric vehicles (EVs) to power networks. Feasibility of the peak-cut of MW-class electricity demands caused by the battery charging of 1000 EVs during late-night is discussed. Ten EV agents connected to a pole-mounted transformer circulate a battery-charing schedule notice among themselves to shrink the excessive electricity demands in their own small group (the interaction among the EV agents belong to the small group). Even though the action of agents is simple and localized, the considerable levelization effect is obtained for the entire system involving the 1000 EVs.

抄録全体を表示

In this paper, multi-agent system is applied into supply chain networks, where multi-item negotiation between multiple manufacturer agents (MA) and multiple material supplier agents (MSA) is discussed. A combined coalition formation protocol is proposed when the order of MA is out of ability of MSA. Not only is coalition formation for complementary considered, but also coalition formation for substitution is taken into account. The coalition which maximizes profit of leader MSA will be determined as the final coalition. Negotiation between MA and final determined coalition is modeled as a Stackelberg game, and the finding of agreement is transformed into determining Stackelberg equilibrium. Simulations are provided to verify feasibility of proposed protocol.

抄録全体を表示

One of important technological issues in microgrid is to control supply and demand balance in own system. In general, the outputs of generators are controlled using the information of frequency which is observed at each generator because the change of the system frequency implies the supply and demand balance in power systems. However, in local systems connected to bulk power system, frequency is not available to control supply and demand balance. In addition, when the owners of DGs are different, it is required that their output should be decided considering economic efficiency. In this paper, to realize such an efficient operation of microgrid, we propose a new supply and demand control method based on mobile agent (MA) technology. In our proposed method, MA intermediates bilateral contracts between customers and DG owners, and negative MA intermediates the negotiation which compensates the output change of renewable energy (RE) by other controllable DGs.

抄録全体を表示

The motion capture systems have been improved, and widely used for motion analysis and synthesis in fields of robotics, animation, rehabilitation, and sports engineering. A lot of captured human data have been accumulated. These prerecorded motion data are expected to be reused. The retrieval of a specified motion data is a fundamental technique for the reuse. This paper describes a novel approach to retrieve motion data from word queries out of a large motion dataset. The motion data are trained by Hidden Markov Models, each of which symbolizes a motion pattern. The motion data are also manually given word labels. The mapping between motion symbols and word labels are optimized through canonical correlation analysis so that correlation between them can be maximized. This mapping makes it possible to project a word query to motion features, and to search for motions similar to the motion features. The validity of the proposed approach was demonstrated on captured motion data.

抄録全体を表示

Recently the research and development concerning electric vehicle (EV) has been progressing. However, the EV system still has problems of battery performance, battery cost, unfinished charge infrastructure, and so on. We have been developing a multi-agent-based traffic simulator to deal with a complex traffic system and to solve various problems. In this paper, we newly implemented EV agents into the traffic simulator. Each EV agent has the formula of precise electricity consumption and the capability to search charge stations. By combining the precise electricity consumption model with the multi-agent-based traffic simulation, we can calculate the cruise distance of each EV in realistic traffic situation. In doing so, we could find the risk of society in near future where EVs become popular. Considering vehicle acceleration and deceleration according to its local traffic situation makes the cruise distance of EVs to have -15% to +15% variation around the mean value. We also analyzed the influence of detour to charging stations.

抄録全体を表示

We built an artificial market model and compared effects of price variation limits, short sell regulations and up-tick rules. In the case without the regulations, the price fell to below a fundamental value when economic bubble bursts occurred. On the other hand, in the case with the regulations, this over-shooting did not occur and the market was more effective. However, the short sell regulation and the up-tick rule caused the trading prices to be higher than the fundamental value. To summarize these points, the price variation limits have the potential to make the market more effective. We also surveyed an adequate limitation price range and an adequate limitation time span for the price variation limit and found a parameters' condition of the price variation limit to prevent the over-shoots.

抄録全体を表示

The purposes of this paper are to build an agent-based model to explain knowledge dynamics in front office and conduct a scenario analysis for evaluating several functions of knowledge acquisition system as one of knowledge management instruments. Conventional agent-based analyses have tackled to some knowledge management problems in organizations. However they did not considered service characteristics when building their model. So we expressly model the characteristics of service co-creation processes between customers and service agents based on some findings in the field of service research. Finally our simulation with the model shows the possible changes of organizational productivity and customer satisfaction by using some scenarios, and then clarifies the reason that why the changes occurred by investigating agents' micro behaviors.

抄録全体を表示

In this paper, we introduce an intelligent pace-car in the traffic flow for the purpose of controlling vehicles that follows a pace-car, and show that it is possible to reduce the phantom traffic jam. Under the situation of phantom traffic jam, the instabilities are observed to grow into traveling waves, which are local peaks of high traffic density, although the average traffic density is still moderate, where the highway is not fully congested. The pace-car manages its velocity to control the following vehicles which are forced to brake when they run into such waves.
The management strategy of the velocity is acquired by reinforcement learning. We employ the extended Nagel-Schreckenberg model which make the traffic flow maximum. By introducing the learned pace-car, we successfully achieves a phase transition that shifts traffic flow from congestion phase to metastable phase.
First, we explain our proposed extended Nagel-Schreckenberg model, and secondly, it is defined that the situation of congestion as the state space to make pace-car learn by reinforcement learning approach. Third, though pace-car finally realizes the highest traffic flow, we evaluate the traffic loss during the period of pace-car's control. As the result. Finally, through the loss evaluation, we show the effectiveness of our approach to acquire the control strategy of pace-car.

抄録全体を表示

The GARCH model is effective to describe the fluctuation process of financial price. It is a model to observe the non-uniform variance of time-series data of financial prices. However, definite theory to reveal how the GARCH effect is produced yet although behavioral characteristics of investors seem to be involved in the causes of GARCH effect. In this study, we aim to reveal the mechanism of producing the GARCH effect in artificial market simulations. We focus on frequency of learning investment strategy.
The simulation yielded the results as follows: the lower the probability of changing investment strategy, the higher the GARCH effect. Results suggested that the GARCH effect relate to the ratio of changing strategies. Another simulation showed that inefficient market can occur the GARCH effect. All simulations same result, these findings could be regarded as general differences, caused by an uniform frequency of learning. The GARCH effect is caused by inefficient markets and ineffective investors.

抄録全体を表示

This paper focuses on the distributed control of the multiple robots which may be broken and investigates how the robots complete their task by collecting broken robots through the simulation of the large-scale structure assembly. For this purpose, we conduct multiagent simulation for collecting broken robots under the different failure rate of robots. Through the intensive simulations, we have revealed that a collection of broken robots before completing their own task (i.e., deploying their panel) is more effective than after complete their own task.

抄録全体を表示

Global warming and a rising energy prices have recently motivated various countries to promote introduction of renewable energy generations. However, energy generations such as photovoltaic power generation bring about unexpected fluctuation and thereby might violate simultaneous balancing electricity demand and supply. Against this situation, the Smart Grid concept has lately attracted attention. In the smart grid concept, mutual information exchange among suppliers and consumers can be achieved to balance and optimize the supply and demand of electricity. In addition, the necessity for electricity trade by which small-scale consumers such as households buy and sell electricity is now advocated to realize further stability of the grid system. This study proposes new trading mechanisms applied in the electricity trade and evaluates them in terms of electricity price and social surplus in the market. All the four mechanisms we propose in this study are simple enough for small-scale consumers to trade electricity everyday and we examine their validity using multi-agent simulations. Results demonstrate that the market-driven mechanism integrating all demand and supply attains higher surplus.

抄録全体を表示

A learning classifier system (LCS) is a model of an intelligent agent interacting with an environment. Many complex yet powerful LCS models exist today. However, they are designed with a single agent approach in mind. LCS applications in multi-agent environment have been problematic. Their complexity limits the agents' cooperation and coordination abilities. This study proposes a simple LCS model for a multi-agent system that allows agents to cooperate and coordinate their actions. New learning methods inspired by organizational learning theories are introduced, giving the agents a capability to recognize useful knowledge. It not only prevents the knowledge from being “forgotten” due to evolutionary process, but also transfers it into less experienced agents. Results show that, with these implementations, the agents manage to coordinate actions better than typical LCS model.

抄録全体を表示

These days the deveopment of information technology has enabled us more easily to get information on user behaviors on the Internet such as web browsing history and products purchase on the E-commerce sites. E-commerce and Web advertising companies try to make effective use of those data to their businesses by segmenting or modeling their customers. In particular, Web advertising companies analyze browsing histories and interesting advertisements of their customers and intend to display ads which their customers likely to have interests in. These changes have influence on marketers, who make decision about ads delivery. They are required to image and decide their targets, or try some delivery and find them. This study introduces the concept of site browsing behavior type and models customer behaviors on browsing and clicking ads on the Internet based on actual customer browsing history data. This study also assumes that this model should be helpful to the marketers, constructs and runs agent simulations with modelized users to validate the user model, and finally analyzes some particular examples and shows the model is useful for customer targeting.

抄録全体を表示

In this study, we propose an agent model based on SECI model by Nonaka for simulating knowledge propagation in organizations. In this model, complexity of knowledge is expressed as bit-tag, and worker-agent and knowledge manager agent are introduced. Some parametric studies is performed for bit-tag length, internalization rates and communication pattern of worker-agent, and evaluate acquirement of knowledge to verify the present simulation model. Furthermore, scenario simulation is implemented with respect to employment periods of worker-agent and the effectiveness of multi-agent model is shown from qualitative evaluation of knowledge variation corresponding to joining and leaving organization.

抄録全体を表示

Recently, office buildings and commercial facilities are getting larger, and emergency evacuation guidance procedures are urgently required. To support evacuation planning, several kinds of evacuation simulations have been proposed. These use walking velocity models that were generated depending on actual pedestrian flow to define an agent's velocity. However, most of these conventional models have been simplified and it is difficult to reproduce complex evacuation scenarios faithfully. In this paper, we propose a walking velocity model for accurate pedestrian simulations. The model presents the relation between pedestrian density and velocity distribution; it was generated through analyzing flows observed from actual evacuation drills. We modeled dense pedestrian flows using the flow data with conventional models to improve simulation performance. In addition, we introduced a method of representing difference among individuals. The validity of the model is confirmed by experimenting with the pedestrian simulator.

抄録全体を表示

It becomes increasingly important for service industries to understand customer behavior using large-scale data such as POS data. However, limitations exist in a customer model constructed on the basis of such behavioral data alone. This paper presents how we can construct a customer model on the basis of both large-scale purchase data and lifestyle survey data. It proposes a method that reveals the connection between lifestyle and behavior by deducing lifestyle from behavioral data using Random Forests, a machine learning algorithm. Then, It applies the proposed method to an actual mass merchandizers using questionnaires on lifestyle collected and the customer behavioral data (ID-POS Data). It thereby demonstrates the effectiveness of the proposed method and its possible use in supporting managerial decision-making on critical issues such as product selection.

抄録全体を表示

As a steady new network communication tool, Social Media has reached global propertion. This phenomenon gave impacts on societies of all over the world. Above all, people provide information to Twitter and Facebook on a daily basis. So that, vast amounts of data exist on Twitter and Facebook. We expect to gather useful information from Twitter and Facebook. Twitter was changed by The 2011 off the Pacific coast of Tohoku Earthquake. And Twitter greatly contributed to diffusion of information. For example, many users checked the safty of their friends or family. However a lot of false rumor were diffused probably because the source of information is unclear. For disasters which will occur, we must analyze diffusion of information on social media as soon as possible. In this paper, we analyze “How the diffusion of information on Twitter has been influenced by structural change of the network which is made by communication among users. ” As the result, just after The 2011 off the Pacific coast of Tohoku Earthquake the network has become easier to diffuse various information. However this means that misinformation are diffused too. We also found that a few users believed misinformation without being corrected.

抄録全体を表示

Connection-strength estimation based on multi-neuron spike data enables us to estimate effective connectivity between neurons and is widely used for investigating information processing in neuronal networks. However, most previous studies have not focused on the networks which partially consist of modulator neurons. We applied connection-strength estimation to categorize neuron types from multi-neuron spike data which were recorded from cultivated neuronal networks including dopamine ones. Substantia nigra and ventral tegmental area cells obtained from rat embryos were cultivated on microelectrode array substrate. Spontaneous electrical activity was observed after 4 weeks in culture. Connection strengths were estimated at the state without antagonists and that with some antagonists. We sorted neurons into 3 neuron types which are assumed as glutamate, GABA, and dopamine ones. Some differences of the activity pattern were observed between each neuron type. These results suggest that neuron type sorting based on connection-strength estimation can be applied to estimate neurons in the networks including dopamine ones.

抄録全体を表示

Serotonin (5-hydroxy tryptamine; 5-HT) plays major roles in mood and cognition. Dysfunction of 5-HT system leads to serious mental illnesses such as major depressive disorder. Little is known, however, about how 5-HT modulates the activity of neuronal network. Here, we evaluated the effects of 5-HT on the activity of cultured neurons obtained from the prefrontal cortex. Dissociated neurons were cultured on a microelectrode array (MEA), and stimulated using 5-HT and selective 5-HT receptor (5-HTR) agonists (8-OH-DPAT and 2-Me-5-HT were used as 5-HTR_1A and 5-HTR₃ agonist, respectively). The bursting activity generated from the cultured neurons was recorded from the electrodes of the MEA. The inter burst intervals (IBI) increased after 5-HT and agonists treatment. Moreover, we evaluated functional connectivity using a delayed transfer entropy method. Pharmacological treatment resulted in enhancement of the connectivity in the entire network. Agonists had apparent effects as the changes of IBI and functional connectivity compared with 5-HT. These results suggest that 5-HT has roles in enhancement of IBI and functional connectivity via 5-HTR_1A and 5-HTR₃.

抄録全体を表示

Accurate temperature control of brain tissue during hypothermia treatment is necessary in order to prevent secondary brain damage and to avoid various side effects. Thus, visualization of intracerebral temperature distribution in hypothermia treatment was basically studied. For this purpose a technology of virtual reality was applied to synthesize a mathematical model that reflects the metabolic heat production and Fourier's heat conduction in a brain with necessary parameters from the various clinical models. In the present study, the experimental system was developed to examine the mathematical simulation concerning with blood flow of a human head by the introduction of brain solid model constructed using silicon rubber of a brain shape based on MRI data, taking into account metabolic heat given by three sheets of film heaters including six sensors for measurement of regional brain temperature. The mathematical simulation describes an internal temperature distribution of brain with similar structure of brain solid model. Results of mathematical simulation and experiment using brain solid model were well consistent in the steady state including control of regional temperature. It enables us to perform experiment of heat conduction in a similar condition of human case, in which inner temperature is clinically difficult to observe. Thus, this mathematical simulation was confirmed useful together with experiment using solid model for the study of future brain hypothermia.

抄録全体を表示

In this paper, a swing-up controller for the Pendubot which is a typical example of nonlinear underactuated systems is proposed. The proposed method consists in three steps, Step 1: a swing-up control of a first link, Step 2: a swing-up control of a second link with stabilizing the first link at upright position, and Step3: a stabilization control of both links at its unstable equilibrium state. As the controller of Step1, a sliding-mode type partial linearization method, which has the robustness for input disturbances, is applied to swing-up and stabilize of the first link. The controller of Step2 is consisted of an energy based swing-up controller for the second link and stabilization controller for the first link designed in Step1. A control input of Step2 is computed by adding inputs of two controllers, and the input of other controller is regarded as a input disturbances,respectively. Using this idea, the swinging-up of the second link with stabilizing the first link is achieved. And sliding-mode controller to stabilize both links is used as the controller of Step3. Finally, numerical simulation and experiment results are given to show an effectiveness of the proposed method.

抄録全体を表示

The technology of face tracking plays an important role for improving the quality of security by surveillance camera and, archiving the smooth communications between human and robot. As one of the face tracking methods, Active Appearance Model (AAM) which is robust for object's rigid and non-rigid transformations has been proposed by T. F. Cootes et al. Although the method is appropriate for the larger face in an image, it is indicated that face tracking for the small face in an image is difficult. Therefore, in this paper, we present the Resolution-variable AAM. Resolution-variable AAM is the novel model for small face fitting in the image which has been difficult to fit to. In the proposed method, we overcome the problem by introducing the structure of changing the resolution to AAM and, changing its resolution adaptively to the resolution of input image. From the simulation results, it was confirmed that the proposed method was rapid and useful for small face fitting.

抄録全体を表示

Proactive maintenance is important to keep normal operation of plant and equipment systems. An equipment monitoring system for such systems was developed that includes following functions: anomaly detection based on threshold method which detects anomalies by comparing a sensor signal with a threshold, anomaly detection based on data-mining which detects anomalies using statistical analysis and rule setting support for the threshold method using the result of the data-mining. The proposed method of rule setting support sets rules for the threshold method in the following procedures. First, training data is generated based on the result of data-mining. Next, a decision tree is generated by learning the training data, and an if-then rule is extracted. Usefulness of the proposed method was evaluated using 4 data sets obtained from real systems. A simple and understandable rule was extracted from a data set. The extracted rule can detect anomalies properly from the other data sets including the same fault. The rule also can explain the reason why the anomalies were detected by data-mining.

抄録全体を表示

We propose a high-speed 3-D object pose recognition method. The method's main feature is that a set of distinctive 3-D vector pairs, each of which consists of three different 3-D points, is used for matching process. A vector pair with a low occurrence probability of a model object means that it is distinctive not only in a model but also in an acquired image. Therefore, such vector pairs are expected to avoid false matching. Experimental results have shown that the proposed method is about 60 times faster and increases the recognition success rate from 62.0% to 94.6% in comparison with the Spin Image method.

抄録全体を表示