Skip to main content
main-content

Über dieses Buch

Multi-Agent System (MAS) is an exciting, emerging paradigm expected to play a key role in many society-changing practices. The International Conference on Principles and Practice of Multi-Agent Systems (PRIMA) is a leading scientific conference for research on intelligent agent systems and multi-agent systems, attracting high quality, state-of-the-art research from all over the world. PRIMA’09 was the 12th in the series of PRIMA conferences and was held in Nagoya, Japan. Beside a single-track main conference, PRIMA’09 also included a number of workshops which were designed to provide a forum for researchers and practitioners to present and exchange the latest developments at the MAS frontier. This book constitutes the post-proceedings of workshops under PRIMA’09. Readers will be able to explore a diverse range of topics and detailed discussions related to a number of important themes in our ever changing world. This collection plays an important role in bridging the gap between MAS theory and practice. It emphasizes the importance of MAS in the research and development of smart power grid systems, decision support systems, optimization and analysis systems for road traffic and markets, environmental monitoring and simulation, and in many other real-world applications and publicizes and extends MAS technology to many domains in this fast moving information age.

Inhaltsverzeichnis

Frontmatter

Agent-Based Collaboration, Coordination and Decision Support

Frontmatter

DGF: Decentralized Group Formation for Task Allocation in Complex Adaptive Systems

Abstract
In this paper, a decentralized group formation algorithm for task allocation in complex adaptive systems is proposed. Compared with current related works, this decentralized algorithm takes system architectures into account and allows not only neighboring agents but also indirect linked agents in the system to help with a task. A system adaptation strategy is also developed for discovering effective system structures for task allocation. Moreover, a set of experiments was conducted to demonstrate the efficiency of our methods.
Dayong Ye, Minjie Zhang, Danny Sutanto

Cellular Automata and Immunity Amplified Stochastic Diffusion Search

Abstract
Nature has often provided the inspiration needed for new computational paradigms and metaphors [1,16]. However natural systems do not exist in isolation and so it is only natural that hybrid approaches be explored. The article examines the interplay between three biologically inspired techniques derived from a plethora of natural phenomena. Cellular automata with their origins in crystalline lattice formation are coupled with the immune system derived clonal selection principle in order to regulate the convergence of the stochastic diffusion search algorithm. Stochastic diffusion search is itself biologically inspired in that it is an inherently multi-agent oriented search algorithm derived from the non-stigmergic tandem calling / running recruitment behaviour of ant species such as Temnothorax albipennis. The paper presents an invesitigation into the role cellular automata of differing complexity classes can play in order to establish a balancing mechanism between exploitation and exploration in the emergent behaviour of the system...
Duncan Coulter, Elizabeth Ehlers

Related Word Extraction Algorithm for Query Expansion – An Evaluation

Abstract
When searching for information a user wants, search engines often return lots of results unintended by the user. Query expansion is a promising approach to solve this problem. In the query expansion research, one of the biggest issues is to generate appropriate keywords representing the user’s intention. The Related Word Extraction Algorithm (RWEA) we proposed extracts such keywords for the query expansion. In this paper, we evaluate the RWEA through several experiments considering the types of queries given by the users. We compare the RWEA, Robertson’s Selection Value (RSV) which is one of the famous relevance feedback methods, and the combination of RWEA and RSV. The results show that as queries become more ambiguous, the advantage of the RWEA becomes higher. From the points of view of query types, the RWEA is appropriate for informational queries and the combined method is for navigational queries. For both query types, RWEA helps to find relevant information.
Tetsuya Oishi, Tsunenori Mine, Ryuzo Hasegawa, Hiroshi Fujita, Miyuki Koshimura

Verification of the Effect of Introducing an Agent in a Prediction Market

Abstract
In recent years, attention to “prediction markets”, which make predictions of the future using market mechanisms, has been increasing. A prediction market applies techniques of experimental markets that have been used in the field of experimental economics to make predictions. A prediction market is a market in which participants make trades of securities predicting the result of a certain event that will be decided in the future. A security provides a dividend based on the result of the event, and the price of the security serves as predictor of the event’s realization probability. A participant predicts the event’s result from various sources of information related to the target phenomenon, and he trades to gain profits from his predictions. From the market price of the result we can suppose think that all the members predictions have been unified. Some of the prediction markets in the U.S. are the Iowa Electronic Market (IEM) and the Hollywood Stock Exchange (HSX). Some of the prediction markets in Japan are at sites such as Shuugi.in and Kounaru. The IEM prediction market in the United States has been effective in predicting election outcomes. It has correctly predicted 75% of the results of elections traded on its exchange, a success rate that compares favorably with that of opinion polls. Thus, in this research, we studied what kind of influence the use of an agent had on a prediction market. For example, we studied how much influence an agent would have on predictive accuracy through an increase in trading volume.
Takuya Yamamoto, Takayuki Ito

A Cognitive Map Network Model

Abstract
This paper presents a cognitive map network model which models causal systems with interactive cognitive maps. Cognitive map is a family of cognitive models that have been widely applied in modeling causal knowledge of various systems like gaming and economic systems. Many causal systems have multiple components which causally evolve concurrently, interacting with each other. Modeling such a system as a whole (cognitive map) is not an ideal solution. Sometimes it is also not possible as the individual parties may not want to release their knowledge to other parties or the coordinating component. The cognitive map network model proposed in this paper represents a causal system as an ecosystem with individual components modeled as cognitive agents. It is a cognitive map ecosystem whose evolution is driven by the component cognitive agents. The cognitive ecosystem model is applied in a toy economic system to illustrate its power in study of hidden patterns.
Yuan Miao

Coordination Strategies and Techniques in Distributed Intelligent Systems – Applications

Abstract
This paper is devoted to describing the broad range of application domains which implement many of the coordination strategies and techniques from the field of multi-agent systems. The domains include defense, transportation, health care, telecommunication and e-business, emergency management, etc. The paper describes the diversity of the applications in which multi-agent coordination techniques have been applied to overcome the challenges or obstacles that have existed with regard to performance, interoperability and/or scalability. While the number of application domains is steadily increasing, the intent of this paper is to provide a small sampling of domains which are applying coordination techniques to build intelligent systems. This paper will also describe an emerging and important problem domain which requires the coordination among many entities across the civil-military boundary, and can benefit from multi-agent coordination techniques.
Abdeslem Boukhtouta, Jean Berger, Ranjeev Mittu, Abdellah Bedrouni

Multi-Agent Area Coverage Using a Single Query Roadmap: A Swarm Intelligence Approach

Abstract
This paper proposes a mechanism for visually covering an area by means of a group of homogeneous reactive agents through a single-query roadmap called Weighted Multi-Agent RRT, WMA-RRT. While the agents do not know about the environment, the roadmap is locally available to them. In accordance with the swarm intelligence principles, the agents are simple autonomous entities, capable of interacting with the environment by obeying some explicit rules and performing the corresponding actions. The interaction between the agents is carried out through an indirect communication mechanism and leads to the emergence of complex behaviors such as multi-agent cooperation and coordination, path planning and environment exploration. This mechanism is reliable in the face of agent failures and can be effectively and easily employed in cluttered environments containing narrow passages. We have implemented and evaluated the algorithm in different domains and the experimental results confirm the performance and robustness of the system.
Ali Nasri Nazif, Alireza Davoodi, Philippe Pasquier

An Approach for Learnable Context-Awareness System by Reflecting User Feedback

Abstract
As the ubiquitous computing becomes popular, the context awareness computing also becomes more interesting reasearch issue. Despite of many important research results on this issue, there still are some limitations that can be enhanced to provide more reliable solution to the user. In this paper, we are describing a design and development of agent based context awareness system that can improve such limitations. The system composed of mainly three layers; the hardware layer receives numerical raw data from sensors and converts it into meaningful semantic data, the middleware layer takes care of ontology modeling, and finally the application layer makes adaptive inference and provides personalized solution to the user. Our approach focuses on the following two main issues. First, we have built ontology based context modeling using fuzzy data that can provide more reliable solution. Second, our CBR based inference engine can provide more personalized and adaptive service by interacting with users feedback. The simulated experimentation has been made and result shows some significant importance.
InWoo Jang, Chong-Woo Woo

A Hybrid Multi-Agent Framework for Load Management in Power Grid Systems

Abstract
In order to cope with load management in power grid systems, this paper presents a hybrid multi-agent framework. This framework integrates the advantages of both centralized and decentralized architectures to achieve both accurate decisions and quick response, and avoid the single point of failure as well. The development of various agents and the behaviors of each agent in the framework are described. Moreover, an example is also introduced, which demonstrates the interaction among agents when a fault happens in a power grid system. The contribution of this paper is to combine local intelligence with global coordination in multi-agent system design to satisfy the challenging requirements in a power grid system.
Minjie Zhang, Dayong Ye, Quan Bai, Danny Sutanto, Kashem Muttaqi

Agent-Based Simulation for Complex Systems: Application to Economics, Finance and Social Sciences

Frontmatter

Financial Frictions and Money-Driven Variability in Velocity

Abstract
Frictions are introduced in the financial structure of a cash-in-advance dynamic stochastic general equilibrium model with the interest of studying their impact on the variability of velocity due to serially correlated monetary shocks. Possessing no analytical solution this dynamic environment, a projection method which parameterizes expectations and employs finite elements in the approximation of the system’s policy functions is executed on the approximation a solution for the equilibrium of the economy and is able to efficiently handle the occasionally binding cash-in-advance constraint on transactions. This last characteristic permits a robust analysis on the impact of frictions on the variability of velocity. It is concluded that frictions on the financial structure of the economy accentuate a precautionary demand for money balances, increasing the incidence of adjustments on the velocity of transactions as an answer to money growth rate shocks.
José J. Cao-Alvira

Automated Fuzzy Bidding Strategy Using Agent’s Attitude and Market Competition

Abstract
This paper designs a novel fuzzy competition and attitude based bidding strategy (FCA-Bid), in which the final best bid is calculated on the basis of the attitude of the bidders and the competition for the goods in the market. The estimation of attitude is based on the bidding item’s attribute assessment, which adapts the fuzzy sets technique to handle uncertainty of the bidding process as well it uses heuristic rules to determine attitude of bidding agents. The bidding strategy also uses and determines competition in the market (based on the two factors i.e. no. of the bidders participating and the total time elapsed for an auction) using Mamdani’s Direct Method. Then the final price of the best bid will be determined based on the assessed attitude and the competition in the market using fuzzy reasoning technique.
Madhu Goyal, Saroj Kaushik, Preetinder Kaur

Resource Allocation Analysis in Perfectly Competitive Virtual Market with Demand Constraints of Consumers

Abstract
Virtual market mechanism solves resource allocation problems by distributing the scheduled resources based on software agent interactions in the market. We formulate agent behaviours negotiating the resource allocations under demand constraints of consumers in the market, and demonstrate the applicability of the virtual market concept to this framework. In this paper we demonstrate the proposed virtual market successfully calculates Pareto optimal solutions in resource allocation problem under the demand constraints of consumers.
Tetsuya Matsuda, Toshiya Kaihara, Nobutada Fujii

Market Participant Estimation by Using Artificial Market

Abstract
In designing a realistic artificial market, one of the most important points to consider is the combination of agents used. In this study, we propose an estimation method based on inverse simulation to estimate the combinations of traders who participate in the market. The proposed method applies a simulation that estimates market paricipation in different markets. The simulation results indicate that the proposed method is capable of estimating market participants.
Fujio Toriumi, Kiyoshi Izumi, Hiroki Matsui

A Study on the Market Impact of Short-Selling Regulation Using Artificial Markets

Abstract
Since the subprime mortgage crisis in the United Sates, stock markets around the world have crashed, revealing their instability. To stem the decline in stock prices, short-selling regulations have been implemented in many markets. However, their effectiveness remains unclear. In this paper, we discuss the effectiveness of short-selling regulation using artificial markets. An artificial market is an agent-based model of financial markets. We constructed an artificial market that allows short-selling and an artificial market with short-selling regulation and have observed the stock prices in both of these markets. We found that the market in which short-selling was allowed was more stable than the market with short-selling regulation, and a bubble emerged in the regulated market. We evaluated the values of assets of agents who used three trading strategies, specifically, these agents were fundamentalists, chartists, and noise traders. The fundamentalists had the best performance among the three types of agents. Finally, we observe the price variations when the market price affects the theoretical price.
Isao Yagi, Takanobu Mizuta, Kiyoshi Izumi

Learning a Pension Investment in Consideration of Liability through Business Game

Abstract
While the importance of financial education is recognized in recent years, the technique for deepening an understanding to pension investment management is needed. In this research, we analyze learning method of the pension investment management in consideration of liability using the business game technique. As a result of analysis, interesting phenomena – the participant understood the learning method of the pension investment management in consideration of liability – were seen. This shows the effectiveness of the business game technique to learning the pension investment management.
Yasuo Yamashita, Hiroshi Takahashi, Takao Terano

An Agent-Based Implementation of the Todaro Model

Abstract
The problem of internal migration and its effect on urban unemployment and underemployment has been the subject of an abundant theoretical literature on economic development. However, most discussions have been largely qualitative and have not provided enough rigorous frame-works with which to analyze the mechanism of labor migration and urban unemployment. In this paper, we build up an economic behavioral model of rural-urban migration which is an agent-based version of the analytical Todaro model described by deterministic ordinary differential equations. The agent-based model allows to explore the rural-urban labor migration process and give quantitative results on the equilibrium proportion of the labor force that is not absorbed by the modern industrial economy.
Nadjia El Saadi, Alassane Bah, Yacine Belarbi

New Types of Metrics for Urban Road Networks Explored with S3: An Agent-Based Simulation Platform

Abstract
The metric of the current road networks tends intrinsically to favour the efficacy of the routes which have the longer range, what leads to promote urban sprawl and automobile dependence. Indeed this metric ensures to individuals the possibility to travel even further, without necessarily increasing their transportation time in the same proportions. According to this assessment, we introduce a new kind of metric, the “slow metric”, which amounts to invert the current ratios of efficacy between the different types of automobile travels, i.e. to favour the efficacy of the short range travels. This metric is reached thanks to traffic lights, under constraints regarding their location and duration. The calibration of the slow metric (number, duration and location of the traffic lights for different networks structures), its impact on traffic (fluidity versus congestion) and conversely the impact of traffic interactions on the slow metric, are simulated and evaluated with S3 which is an agent-based simulation platform.
Cyrille Genre-Grandpierre, Arnaud Banos

Impact of Tenacity upon the Behaviors of Social Actors

Abstract
The Sociology of the Organized Actions is a well-established theory that focuses upon the actual behaviors of the members of social organizations, and reveals the (to a large extent implicit) motives of social actors. The formalization of this theory leads to model the structure of an organization as a social game, including the Prisoners’ Dilemma as a specific case. In order to perform simulations of social organizations modeled in this way, the SocLab environment contains an algorithm allowing the model’s actors to play the social game and so to determine how they could cooperate with each other. This algorithm includes several parameters, and we study the influence of one of them, the Tenacity.
Joseph El-Gemayel, Christophe Sibertin-Blanc, Paul Chapron

Agent Technology for Environmental Monitoring and Disaster Management

Frontmatter

Dynamic Role Assignment for Large-Scale Multi-Agent Robotic Systems

Abstract
In this paper, we introduce an approach for designing and deploying organizations on large scale Multi-Agent Robotic Systems. Lying on the well-known organizational meta-model AGR, we propose a role assignment protocol for automatically distributing the roles over the robots. We have run a series of simulations to validate the approach’s feasibility. The simulations show that the protocol is well scalable as well.
Van Tuan Le, Serge Stinckwich, Bouraqadi Noury, Arnaud Doniec

Multi-agent System for Blackout Prevention by Means of Computer Simulations

Abstract
The paper presents a dynamic simulation model of power flows in a power distribution network using communication in multi-agent systems. The model is based on local interchange of knowledge between autonomous agents representing the network elements. The agents use KQML as an inter-agent knowledge interchange language. The main purpose of the work is to develop a scalable distributed simulation model, flexible enough for incorporation of intelligent control and condition-based management of a power distribution network. This model is designed to optimize the electric distribution networks and prevent failures in the power grid (blackout). The network’s reliability is affected by the parameters of connected power sources (both traditional and renewable sources), but also by the unexpected failures. The simulation is using the parameters calculated from the real database of electric distribution network failures.
Miroslav Prýmek, Aleš Horák, Adam Rambousek

Asking for Help through Adaptable Autonomy in Robotic Search and Rescue

Abstract
Robotic search and rescue teams of the future will consist of both robots and human operators. Operators are utilized for identifying victims, by means of camera feeds from the robot, and for helping with navigation when autonomy is insufficient. As the size of these robot teams increases, the mental workload on operators increases, and robots find themselves in precarious situations with no assistance for resolution. This paper presents an approach that utilizes multiple levels of autonomy to allow a robot to consider a range of options, including asking for operator assistance, for dealing with problematic situations to maximize efficient use of the operator’s time. Individual robots use self monitoring to determine failures in task progression, a form of local autonomy. Upon this trigger, the robot evaluates decisions to properly route asking for help, consensus autonomy. A Call Center alerts the operator(s) to incoming requests for assistance. This results in a better use of operator time by focusing attention where it is needed. Experiments explore the effectiveness of agents’ decisions, both local and team-level, with multiple simulators. A high fidelity simulator and user interface further evaluate how effective robot information is relayed to the operator, through human trials.
Breelyn Kane, Prasanna Velagapudi, Paul Scerri

Conceptual Framework for Design of Service Negotiation in Disaster Management Applications

Abstract
Interactions between stakeholders providing their services in disaster management might involve negotiations for optimal selection of service providers. In this paper we propose a conceptual framework for designing cooperative multi-issue one-to-many service negotiations. The framework allows us to define: negotiation protocols, negotiation subject, properties of negotiation subject issues, deal spaces, and utility functions of participant agents. We also consider a simple example of how this framework can be instantiated to particular negotiation in a disaster management information system.
Costin Bădică, Mihnea Scafeş

An Optimized Solution for Multi-agent Coordination Using Integrated GA-Fuzzy Approach in Rescue Simulation Environment

Abstract
Agents’ coordination, communication and information sharing have been always open problems in multi-agent research fields. In complex rescue simulation environment, each agent observes a large amount of data which exponentially increases through the time while the capacity of messages in which agents’ information is shared with others and also the time needed to process the data is limited. Apparently, having a suitable coordination strategy and data sharing system will lead to better overall performance. Therefore, agents should select and spread out the most useful data among their observed information in order to achieve better coordination. In this paper, we propose an unique method based on integration of Genetic Algorithms and Fuzzy Logic theory to decide which part of data is more important to share in different situations. We also advise a new iterative method in order to obtain admissible experimental results in rescue simulation environment which is a good measurement for our research.
Mohammad Goodarzi, Ashkan Radmand, Eslam Nazemi

A Study of Map Data Influence on Disaster and Rescue Simulation’s Results

Abstract
Applying the agent-based simulation system to real cases requires the fidelity of the simulations. It is desired to simulate disasters situations under environments that are close to real ones qualitatively and quantitatively. Maps are a key component of the disaster and rescue simulation system. We propose a method to create more real maps using open GIS data and discuss the simulation results between two maps. Buildings data of one map is generated by programs and the other is created from real data. The experiments show clear difference between the simulation results, and the causes of the difference are discussed.
Kei Sato, Tomoichi Takahashi

Evaluation of Time Delay of Coping Behaviors with Evacuation Simulator

Abstract
As described in this paper, we analyzed the influence of the time necessary to begin coping behaviors on the damage caused by chemical terrorism. To calculate the damage of a chemical attack in a major rail station, our network model-based pedestrian simulator was applied with systems designed to predict hazards of indoor gas diffusion. Our network model is designed to conduct simulations much faster, taking less than few minutes for simulation with ten thousands of evacuators. With our evacuation planning assist system, we investigated a simulated chemical attack on a major rail station. In our simulation, we showed a relation between the time necessary to begin coping behaviors of the managers and the damage to passengers. Results of our analyses were used for the instruction of rail station managers in a tabletop exercise held by the Kitakyushu City Fire and Disaster Management Department.
Tomohisa Yamashita, Shunsuke Soeda, Itsuki Noda

Web-Based Sensor Network with Flexible Management by an Agent System

Abstract
This paper presents the architecture and function of an agent system developed for sensor nodes equipped with a Web server such as Field Servers to manage a Web-based sensor network. The agent system accesses flexibly all Web components with a rule-based function and it performs various types of monitoring and complicated operations. By accessing useful Web applications such as image processing, the system also provides versatile data analysis. Moreover, the agent system has Web interfaces to control itself on the Web, so we can construct a multi-agent system in which one agent system controls not only Field Servers but also other agent systems to achieve high scalability and a robust system.
Tokihiro Fukatsu, Masayuki Hirafuji, Takuji Kiura

Sensor Network Architecture Based on Web and Agent for Long-Term Sustainable Observation in Open Fields

Abstract
Architecture of sustainable, robust and scalable sensor networks is proposed. To monitor environmental condition and ecosystems for many years in open fields, Web-based Sensor network can be employed. Agents control them by using purely HTTP. Functions of the sensor nodes can be reduced as much as the agent can assist the sensor nodes. Applications for the sensor networks are distributed Web services. All devices in the sensor nodes are Web servers such as data acquisition Web-server card and IP camera, and they are connected by Ethernet. That is, all devices are only Web servers. Power supplies for all devices in the sensor nodes are controlled by the agent, which can restart the freezing devices automatically. The Web-based sensor networks were tested at many sites, and rationality of the proposed architecture was examined.
Masayuki Hirafuji, Tokihiro Fukatsu, Takuji Kiura, Haoming Hu, Hideo Yoichi, Kei Tanaka, Yugo Miki, Seishi Ninomiya

A Multi-Agent View of the Sensor Web

Abstract
The rapid growth in sensor and ubiquitous device deployments has resulted in an explosion in the availability of data. The concept of the Sensor Web has provided a web-based information sharing platform to allow different organisations to share their sensor offerings. We compare the Open Geospatial Consortium - Sensor Web Enablement (OGC-SWE) with Multi-Agent System (MAS), and identify the similarities between the concepts. These similarities motivate the adoption of MAS based techniques to address related problems in OGC-SWE. Brokerage facilitators commonly used in MAS, the Yellow Pages Agent and Blackboard Agent, are considered to address service interaction issues identified within OGC-SWE. Furthermore, the use of MAS based reputation mechanisms are explored to address potential trust issues between service providers and consumers in OGC-SWE.
Quan Bai, Siddeswara Mayura Guru, Daniel Smith, Qing Liu, Andrew Terhorst

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise