Principles of Practice in Multi-Agent Systems

Electronic commerce has been a significant commercial phenomenon in recent years and autonomous agents have made the advantages of e-markets more distinct. However, as e-market environments become open and dynamic, existing agent negotiation approaches expose some limitations. Static negotiation strategies and offer evaluation approaches might fail to capture dynamic changes of market situations, as well as changes of negotiators’ expectations on negotiation outcomes. When market situations change, agents may need modify their negotiation strategies, expectations and criteria on offer evaluations as well as counter-offer generations in order to maximize their profits. Furthermore, in multi-issue negotiations, agents may have multiple preferences, which might not be delivered by most of existing negotiation approaches. In this paper, we propose a market-based multi-issue negotiation model to capture the dynamic changes of negotiation environments and impacts on negotiation strategies, counter-offer generations and offer evaluations. Also, the proposed model allows negotiators to deliver multiple offers to match their different preferences and negotiators would have more chances to reach agreements. Experimental results illustrate improvements of the proposed model on negotiators’ utilities and efficiencies of the whole negotiation system by comparing with the performance of NDF negotiation model.

In this paper, we present a protocol for collaborative translation, where two non-bilingual people who use different languages collaborate to perform the task of translation using machine translation (MT) services. Members in one real life example of intercultural collaboration try to share information more effectively by modifying unnatural machine translated sentences manually and improving their fluency. However, there are two problems with this method: One is that poor quality of translation can induce misinterpretations, and the other is that phrases in the machine translated sentence that a person cannot make sense of remain unmodified. The proposed protocol is designed to solve these problems. More concretely, one person, who handles the source language and knows the original sentence (source language side), evaluates the adequacy between the original sentence and the translation of the sentence modified to be fluent by the other person, who handles the target language (target language side). In addition, by determining whether the meaning of the machine translated sentence is understandable, it is ensured that the two non-bilingual people do above tasks properly. As a result, this protocol 1) improves MT quality; and 2) terminates successfully only when the translation result becomes adequate and fluent. The experiment results show that when the protocol terminates successfully, the quality of the translation increases to about 83 percent in Japanese-English translation and 91 percent in Japanese-Chinese translation.

There is a growing belief that the environment plays an important role in the healing process of patients, supported by empirical findings. Previous research showed that psychological stress caused by loneliness can be reduced by artificial companions. As a pilot application for this purpose, this paper presents an affective agent playing tic-tac-toe with the user. Experimenting with a number of agents under different parameter settings shows the agent is able to show human-like emotional behavior, and can make decisions based on rationality as well as on affective influences. After discussing the application with clinical experts and making improvements where needed, the application can be tested in a clinical setting in future research.

To avoid the occurrence of spirals of negative emotion in their teams, team leaders may benefit from intelligent agent systems that analyze the emotional dynamics of the team members. As a first step in developing such agents, this paper uses an agent-based approach to formalize and simulate emotion contagion spirals within groups. The computational multi-agent model is integrated within an intelligent ambient agent to monitor and predict group emotion levels over time and propose group support actions based on that.

In this paper, we proposed a statistical utterance selection method for dialogue agents by applying a machine learning algorithm. We defined statistical candidate utterance selection as a question that automatically selects an appropriate utterance from speech collections prepared in advance as responses. To realize automatic utterance evaluation, we employed manually evaluated data as learning data so that relative magnitude correlation will be learned from them.

We checked the order of the automatically evaluated values to prove the validity of our proposed method. In this simulation, the result shows that the top appropriate utterance is selected at 47.5%, and it is selected within the top 10 at 78.0%. For implementing this method in agents that assist humans by replying, we found that it is quite possible to realize such an agent.

In ways analogous to humans, autonomous agents require trust and reputation concepts in order to identify communities of agents with which to interact reliably. This paper defines a class of attacks called witness-based collusion attacks designed to exploit trust and reputation models. Empirical results demonstrate that unidimensional trust models are vulnerable to witness-based collusion attacks in ways independent multidimensional trust models are not. This paper analyzes the impact of the proportion of witness-based colluding agents on the society. Furthermore, it demonstrates that here is a need for witness interaction trust to detect colluding agents in addition to the need for direct interaction trust to detect malicious agents. By proposing a set of policies, the paper demonstrates how learning agents can decrease the level of encounter risk in a witness-based collusive society.

In this paper, we present an automated multi-agent multi-issue negotiation solution to solve a resource allocation problem. We use a multilateral negotiation model, by which three agents bid sequentially in consecutive rounds till some deadline. Two issues are bundled and negotiated concurrently, so win-win opportunities can be generated as trade-offs exist between issues. We develop negotiation strategies of the agents under an incomplete information setting. The strategies are composed of a Pareto-optimal-search method and concession strategies. An important technical contribution of this paper lies in the development of the Pareto-optimal-search method for three-agent multilateral negotiation. Moreover, we present the identification of agreements and Pareto-optimal outcomes achieved by our methods in mathematical proof. We show through computer experiments that using the tractable heuristic of Pareto-optimal-search combined with well-designed concession strategies by agents results in (near) Pareto-optimal outcomes.

In a cooperative heterogeneous multiagent team, distributed agents are required to harmonize activities and make the best use of resources to achieve their common goal. Agents are required to share their resource with very a few of the teammates who need it but with a limited view of the team, they do not know who they are. In this paper, we put forward our resource sharing algorithm for a large heterogeneous team. It does not require a complete view of the team or depend on excessive communication. Agents only make use of the knowledge from allocating tasks or sharing the other resources. The key is that we use influence diagram to model how agents may predict what the other agents are doing from their limited information received. By utilizing the relevances between tasks and resources or pairs of resources, We have setup a local probability model so that agents can reason in the uncertainty and can efficiently share the resource within a few hops to its target. Based on this model, we have two additional designs of dynamic threshold and local decision exchange model so that agents can enhance their local decisions and greatly increase the resource sharing performance. Our experiment results show this system design is feasible for resource sharing in a large heterogeneous multiagent team.

Gaia is a well-known Agent Oriented Software Engineering (AOSE) methodology. The emerging Model-Driven Engineering (MDE) paradigm encourages software modelers to automate the transition of one type of software model to another and eventually the code generation process. Towards this end we define a process for transforming the Gaia roles model liveness formulas to statecharts. This achievement on one hand allows the modeler to work on detailed agent design and permits, on the other hand, to automatically generate an agent’s code using any one of the statecharts-based tools in the market.

In the research area of web technologies, ontologies are recently widely used. By using ontologies, we can share common understanding of the structure of information among people or software agents and enable reuse of domain knowledge. However, the difficulties in building ontologies have been pointed out and its costs are raising problems currently. To build an ontology, we must determine the domain that the ontology will cover, and define taxonomy, properties, instances of the ontology. It is very difficult and time consuming to build them without any tools. In this paper, we propose ONTOMO that enables Internet users to take part in building ontologies as a part of collective intelligence. In particular, we present an instance recommendation mechanism based on the editing history of multiple users together with experimental evaluations.

In this paper, we attempt to formalize the syncretic argumentation, taking into account the Golden Rule in the ethics of reciprocity and Confucius’ Golden Rule. After outlining the underlying argumentation framework, Logic of Multiple-valued Argumentation (LMA), we describe the syncretic argumentation framework by introducing the lattice homomorphism on truth-values (epistemic states) of propositions, and the new definitions of arguments justified under syncretized knowledge base. We also argue about its implications and new directions to further work.

Software agents can act flexibly in a variety of electronic marketplaces. Continuous Double Auction (CDA) is an efficient and common form of these marketplaces. There are several bidding strategies proposed in the literature for agents to adopt to compute their asks or bids in CDAs. For all of these bidding strategies, starting price has not been taken into account. However, in online auction marketplaces, the starting price is an important parameter for sellers to set and has been discussed many a time in the literature. Given the importance of starting price, the main objective of our work is to explore the effect of starting price on agents using various bidding strategies and how to adjust it adaptively within a dynamic CDA market. Experimental results confirm that when agents set their starting prices at varying values in different market situations, their profit changes significantly no matter which strategy they adopt. In order to guide agents to adjust their starting prices in dynamic and unknown markets, an adaptive mechanism is proposed. Experimental results show that agents adopting the adaptive mechanism generally outperform the corresponding agents without. Furthermore, another set of experiments are carried out to let all the agents use the adaptive mechanism and compete together in one market. Not surprisingly, the profit of agents is observed to drop down a lot in this situation.

This work presents an agent based system devoted to the simulation of passenger transportation scenarios. The architecture is build over a system, called MADARP, devoted to the implementation of concrete passenger transportation planning Systems. The transportation type considered by the system is the demand-responsive one, that is, a flexible approach in which trips requests are tacked online and scheduled over a set of available vehicles. The simulator allows diverse scenarios by varying the geographical network, the requests, and the set of vehicles. By managing diverse eventualities, it gives dynamicity to the simulation, such as, delays of vehicles, clients’ no-show and vehicles’ breakdowns, among others. The general design is depicted using the PASSI methodology, together with its implementation over the Jade agent platform.

Agent-oriented programming has been motivated in part by the conception that high-level programming constructs based on common sense notions such as beliefs and goals provide appropriate abstraction tools to develop autonomous software. Various agent programming languages and frameworks have been developed by now, but no systematic study has been done as to how the language constructs in these languages may and are in fact used in practice. Performing a study of these aspects may contribute to the design of best practices or programming guidelines for agent programming, and clarify the use of common sense notions in agent programs. In this paper, we analyze various agent programs for

dynamic blocks world

, written in the

Goal

agent programming language. We present several observations based on a quantitative and qualitative analysis that provide insight into more practical aspects of the development of agent programs. Finally, we identify important issues in three key areas related to agent-oriented programming that need further investigation.

Service-Oriented Architectures (SOA) provide infrastructures to make resources available to other participants in the network as independent services. However, service providers, not having the autonomy to decide who they collaborate with, might be reluctant to participate in such open systems, the client being the sole responsible of the selection of services for the composition. Multiagent systems research offer some solutions in term of trust and reputation mechanisms as well as in coalition formation theory. This paper presents a multiagent based negotiation model to enable provider autonomy in composite web services. QoS-based reputation is built from both feedbacks retrieved from execution and from subjective feedbacks given by the client. This model is illustrated by an example based on the Language Grid Project, an service infrastructure for language resources.

Automated Trust Negotiation(ATN) has been proposed as a mechanism to establish mutual trust among strangers. Protocols and strategies to be used during ATN have also been studied. When considering the real world usage of ATN, there are many factors to be considered. One of the factors that has not been addressed by previous studies is the memory complexity of negotiation strategies. This paper analyses the memory complexities of previously proposed negotiation strategies and evaluates the average memory consumption through simulations using an ATN framework for web services. The experimental results revealed that memory complexity of Parsimonious strategy grows exponentially as the number of credentials increases, which is consistent with the theoretical analysis. As a solution, a method to reduce the memory consumption by exploiting the knowledge each entity has about the negotiation is presented. In addition, the paper presents a new criterion that enables the truncation of the negotiation to reduce the memory consumption in situations where the negotiation fails. Experiment results, which show the effectiveness of above methods in reducing the memory consumption, negotiation length are also presented.

Distributed sensor network is an important research area of multi-agent systems. We focus on a type of distributed sensor network systems that cooperatively observe multiple targets with multiple autonomous sensors that can control their own view. The problem of allocating observation resource of the distributed sensor network can be formalized as distributed constraint optimization problems. However, in the previous works, the computation cost to solve the resource allocation problem highly increases with its scale/density. In this work, we divide the problem into two layers of problems, and two layered cooperative solvers are applied to those problems. The result of the experiment shows that our proposed method reduces the number of message cycles.

Most real-world negotiations involve multiple, interdependent issues or attributes. These negotiation scenarios are specially challenging because agents’ preferences on the attributes may be non-monotonic. The existing works in the area of non-monotonic preference scenarios are mainly restricted to mediated approaches and to very specific agents’ preference models. In this research we propose NegoExplorer, a generic framework for non-mediated automated bilateral multi-attribute negotiations. NegoExplorer is based on a region-based recursive bargaining mechanism. The mechanism is named recursive because agents negotiate on regions and not on specific contracts in the negotiation space, and because an agreement on a region implies a new bargaining which is constrained to that region. Agents may reach an agreement on a contract by iteratively applying this recursive mechanism. In order to evaluate the effectiveness of our proposal we have compared it with a classical similarity based negotiation protocol. The preliminary results are promising, showing better results in terms of utility and negotiation time for the case of non-monotonic and non-differentiable utility spaces, and similar results for monotonic spaces. We believe that the ideas presented in this paper may be the starting point of a new family of negotiation mechanisms.

This paper proposes a novel Peer-to-Peer Information Retrieval (P2PIR) method using user feedback and query-learning. The method actively utilizes negative feedback information so that other agents can filter it out when retrieving it. The proposed method effectively increases retrieval accuracy and decreases communication loads required for document retrieval in communities.

The experiments were carried out on multiple communities constructed with multi-agent framework

Kodama

[1]. The experimental results illustrated the validity of our proposed method.

This paper presents a human-aware software agent to support a human performing a task that demands substantial amounts of attention. The agent obtains human awareness in an adaptive manner by use of a dynamical model of human attention which is parameterised for specific characteristics of the human. The agent uses a built-in adaptation model to adapt on the fly the values of these parameters to the personal characteristics of the human. The software agent has been implemented in a component-based manner within the Adobe

®

Flex

®

environment.

We have developed a new two-sided matching protocol including job applicants and employers in the condition that applicants have conditional preferences and well informed applicants exist. In past research, two-sided matching has covered some assignment problems such as residency matching. However, in the case of matching on the information network, different applicants are differently informed and well informed applicants hide its information to obtain more desirable matching. That is, asymmetric information possessed by applicants causes unstable matching. To overcome this difficulty, we design a new two-sided matching protocol in which applicants are allowed to report their conditional preferences and well informed applicants generally have an incentive to share information among applicants by allowing applicants to report their conditional preferences and deciding the matching on the basis of the preferences of applicants who share information (informers). We experimentally evaluated our protocol through simulation and found that the protocol can attain more satisfactory matching.

A set of physical feature values, called a Laban’s feature set, is proposed to explain an observer’s impression of bodily expression. The design concept of our Laban’s feature value set is based on Laban Movement Analysis, which is a well known theory in body movement psychology. For practical application to human-agent interaction (HAI), use a human-form robot (HFR) as a motioning object. A correlation between our Laban’s feature set and the HFR’s emotions (

Pleasure

,

Anger

,

Sadness

and

Relaxed

), which subjects estimated, was examined. In reference to Russell’s circumplex model of affect, we discuss the correlation using two axial (“pleasure-displeasure” and “degree of arousal”) characteristics. Finally, we present four estimation equations with accuracy rates of more than 85%.

Internet bots have been widely used for various beneficial and malicious activities on the web. In this paper we provide new insights into a new kind of bot termed as

web spambot

which is primarily used for spreading spam content on the web. To gain insights into web spambots, we developed a tool (HoneySpam 2.0) to track their behaviour. This paper presents two main contributions, firstly it describes the design of HoneySpam 2.0 and secondly we outline the experimental results that characterise web spambot behaviour. By profiling web spambots, we provide the foundation for identifying such bots and preventing and filtering web spam content.

Service-oriented Open technology is becoming more and more the enabler tool for today open enterprise ISs. In this field, it seems interesting to work with the Multi-Agent System research focusing on open distributed, complex and dynamic systems, in which information and resources are distributed among several agents, which interact by means of services. In this paper, an engineering tool for Service-oriented Open Multi-Agent Systems is presented. This tool is based on a platform independent unified meta-model. In this way, a Model Driven Architecture mechanism is applied, thus defining a Service-Oriented Multi-agent System meta-model based on Virtual Organizations and employing the Eclipse technology to develop the IDE tool.

The evaluation of multiagent system software engineering techniques is an open research topic. Nowadays, on the market there are a great number of methods and frameworks to develop multiagent systems. It makes difficult the selection between one and another. In this paper, an evaluation framework for analyzing, comparing and selecting methods and tools for developing multiagent systems is presented.

We aimed to achieve smooth human-robot communication by using a communication model based on synchronization between human and robot paralanguages. We also described a robot’s mental rhythm that controls its own paralanguage and entrains its mental rhythm into a human paralanguage rhythm for human-robot communication. We built three robots to evaluate our proposed model: the first used our communication model and the other two used extreme models that either completely imitated human paralanguage or did not imitate it at all. We prepared several conversations between subjects and each of the three robots. The experimental results revealed that synchronized conversation using human and robot paralanguages gave humans a positive impression of the robots. This paper also reports the results from analyzing the correlation between human and robot paralanguages.

A method of generating new motions associatively from novel trajectories that the robot receives is described. The associative motion generation system is composed of two neural networks: nonlinear principal component analysis (NLPCA) and Jordan recurrent neural network (JRNN). First, these networks learn the relationship between a trajectory and a motion using training data. Second,

associative values

are extracted for associating a new corresponding motion from a new trajectory using NLPCA. Finally, a new motion is generated through calculation by JRNN using the

associative values

. Experimental results demonstrated that our method enabled a humanoid robot, KHR-2HV, to associatively generate the new motions corresponding to trajectories that the robot had not learned.

In this paper, we present and demonstrate

SmartContractor

: a distributed task assignment system based on the simple ContractNet protocol. In Japan, housewives doing side work in their free time at home is common practice which helps those that have children and the ability to do work. Recently, a franchised sideline business has grown. They started by establishing a systematic business process for sideline work shops, then franchising and networking those shops .In such franchised sideline businesses, it is of utmost necessity to allocate or delegate tasks among the shops by using a web-based network system. This business process of task allocation/delegation is very similar to the famous ContractNet protocol. Thus, a ContractNet approach is a smart and straightforward for us. The initial

SmartContractor

system has been delivered, and is improved with an Agile-based development process. The system will be implemented from September 2009.

Network games are attracting attention as simulation platforms for social experiments because of their rich visualization performance and scalability. Our objective in this study is to develop a participatory simulation platform on a network game. Unlike non player characters (NPCs) in network games, agents in a participatory multiagent-based simulation (PMAS) should behave as real-world humans according to behavior models. We developed a novel networked participatory simulation platform called

gumonji

/

Q

by integrating scenario description language

Q

with the network game

gumonji

. This paper details the implementation of

gumonji

/

Q

. In order to connect

Q

and

gumonji

, we implement communication sub-components that realize TCP/IP communication between them, and a scenario translator to convert a request from

Q

into a sequence of operators. This makes it possible for the

gumonji

simulator to deal with human-controlled avatars and

Q

-controlled agents in a unified way.

A more promising technology to integrate existing software systems and their functionalities and to add assistant systems for the shop floors is to be found in software agents. Concerning with the applications of agent technology to intelligent process automation systems, a flexible and extensible approach based on multi-agent system (MAS) and distributed planning technique to reach for increased flexibility and fault-tolerance in process monitoring and control operations is proposed. The monitoring operations are aimed at combining information from different sources depending on the monitoring tasks. The control operations are supervisory control tasks performing either in both sequential and iterative forms. The agent layer is used for monitoring the operations of the lower-level automation systems and reconfiguring its control logic. It operates as a distributed planning and plan execution system to increase the operational flexibility of the whole process automation systems. The software architecture and its functionalities and components for the implementation of the proposed approach are also presented. The design strategy and the discussion on the proposed approach are provided.

In this work a simulation model is described and implemented, with the purpose of analyzing the non-equity collaborations among small and medium enterprises (SMEs) and the effects of innovation management strategies on enterprise networks. Non-equity links are usually stable, but not strong. In this context the strong links are joint-ventures and participation exchanges, while non-equity collaboration (as a consortium) are stable, but leaving each enterprise as an autonomous entity. In particular, the governance of SMEs remains independent, but in the long term we observe a co-evolution of strategies among the enterprises which take part in the collaborative network. An enterprise can decide to exploit innovative processes it owns, thus potentially gaining competitive advantage, but risking, in turn, that other players could reach the same technological levels. Or it could decide to share it, in exchange for other competencies or money. These strategies could be the basis for a network formation and/or impact the topology of an existing network. The model presented in the paper aims at exploring how a process innovation and the strategies to manage it can facilitate network formation, affect its topology, induce new players to enter the market and spread onto the network by being shared or developed by new players.

Multiagent-based simulations are a key part of several research fields. Multiagent-based simulations yield multiagent societies that well reproduce human societies, and so are seen as an excellent tool for analyzing the real world. A multiagent-based simulation allows crowd behavior to emerge through interactions among agents where each agent is affected by the emerging crowd behavior. The interaction between microscopic and macroscopic behaviors has long been considered an important issue, termed the “micro-macro problem”, in the field of sociology, but research on the issue is still premature in the engineering domain. We are focusing on citywide traffic as a target problem and are attempting to realize mega-scale multiagent-based traffic simulations. While macro-level simulations are popular in the traffic domain, it has been recognized that micro-level analysis is also beneficial. However, there is no software platform that can realize analyses based on both micro and macro viewpoints due to implementation difficulties. In this paper, we propose a traffic simulation platform that can execute citywide traffic simulations that include driving behavior models. Our simulation platform enables the introduction of individual behavior models while still retaining scalability.

There is a steady increase of number of people who are suffering from lifestyle-related diseases. Although much work has been done on healthcare support system, these existing systems are limited in ability of healthcare support service. This paper proposes an agent-based framework for advanced healthcare support system. In order to provide useful information for healthcare of an object person, not only to him/herself but also to the related people of that person, the system needs to acquire variety of information, knowledge, data, etc. and store/manage them in a systematic manner. This paper mainly focuses on the concept and design of the system, and also we describe the implementation details of several functions for the healthcare.

Interpolation system of traffic condition is proposed, which consists of estimation and learning agents. To evaluate the interpolation accuracy, coefficient of determination (CD) and mean square error (MSE) are used. The interpolation accuracy can be improved by the alternate use of estimation and learning agents, and the iterative uses of the same probe data. The standard deviation of the normalized velocity can be improved to 0.1353, and that of the velocity is 6.77 km/h in the mid velocity region. Furthermore, the CD and MSE could be improved by the additional repetition of estimation and learning.

Ontologies are a crucial tool for formally specifying the vocabulary and the concepts of agent platforms, so, to share information, agents that use different vocabularies must be able to translate data from one ontological framework to another. The treatment of uncertainty plays a key role in the ontology mapping, as the degree of overlapping between concepts can not be represented logically. This paper aims to provide mechanisms to support experts in the first steps of the ontology mapping process using fuzzy logic techniques to determine the similarity between concepts from different ontologies. For each pair of concepts, two types of similarity are calculated: the first using the Jaccard coefficient, based on relevant documents taken from the web, and the second based on the linguistic relationship of concepts. Finally, the similarity is calculated through a fuzzy rule-based system. The ideas presented in this work are validated using two real-world ontologies.

The intrinsic openness and dynamism of the Service-Oriented Computing (SOC) vision makes it crucial to locate useful services and recognize them as trustworthy. To address this challenge, referral networks have recently been proposed as a decentralized approach based on software agents technology. Although in theory this idea might look promising for enabling the SOC vision, real-world referral systems are still missing. In this paper we study this gap between theory and practice from the point of view of agent communication, since it represents a key feature of agent-based distributed systems. To do this, we firstly highlight the main agent communication requirements needed to cope with real-life agent-based referral networks. Secondly, we discuss why the standard language for agent communication (FIPA ACL) is not suitable for supporting these requirements. Finally, we briefly illustrate how they can be easily satisfied by an advanced agent communication language, namely FT-ACL.

This paper presents SADE, a software supporting environment for developing and running self-adaptive multi-agent systems (MAS). SADE consists of four parts: an adaptive mechanism, a programming language SADL, a reusable software package and supporting software tools. The adaptive mechanism is based on the organization metaphor to analyze and implement self-adaptation of MAS. In our approach, self-adaptation of agent is realized as the changes of roles that agent plays in MAS organization by executing four atomic adaptation operations: “join”, “quit”, “activate” and “deactivate”. SADL is presented to describe the adaptive strategies that express how agents in MAS adapt to the changes of the situated environment. It enables developers to describe self-adaptation explicitly and separate the functional behaviors from adaptation behaviors of agents, thus simplifying the development and maintenance of complex adaptive MAS. SADE also provides a reusable software package that encapsulates the elementary functionalities of self-adaptation, such as the adaptive mechanism, etc. In order to support the development, deployment and execution of adaptive MAS, a compiler and editor for SADL, the architecture of self-adaptive agent and its execution engine have been developed. The technical details of SADE are introduced and a case is studied to illustrate our approach.

In this article, we propose a method to adapt stepsize parameters used in reinforcement learning for non-stationary environments. When the environment is non-stationary, the learning agent must adapt learning parameters like stepsize to the changes of environment through continuous learning. We show several theorems on higher-order derivatives of exponential moving average, which is a base schema of major reinforcement learning methods, using stepsize parameters. We also derive a systematic mechanism to calculate these derivatives in a recursive manner. Based on it, we construct a precise and flexible adaptation method for the stepsize parameter in order to maximize a certain criterion. The proposed method is also validated by several experimental results.

We present a new insurance system for healthcare reform to meet the medical demand and alleviate the cost burden in China. China healthcare reform is complex where unlike most countries’ uniform system; it has two branches: urban health insurance and new rural cooperative medical. The equity and efficiency of the two medical healthcare systems are discussed in this paper. We use multi-agent based computational mechanism design simulation to analyze the healthcare insurance’s coverage, service and treatment cost of the people. A summary of the recent medical healthcare reforms undertaken in China is also discussed. Research results indicate that our novel hybrid healthcare insurance system formed by merging parts of the two branches can improve equity without compromising efficiency.

With the rapid development of case-based reasoning (CBR) techniques, CBR has been widely applied to real-world applications such as agent-based systems for ship collision avoidance. A successful CBR-based system relies on a high-quality case base. Automated case creation technique is highly demanded. In this paper, we propose an automated case learning method for CBR-based agent systems. Building on techniques from CBR and natural language processing, we developed a method for learning cases from maritime affair records. After reviewing the developed agent-based systems for ship collision avoidance, we present the proposed framework and the experiments conducted in case generation. The experimental results show the usefulness and applicability of case learning approach for generating cases from the historic maritime affair records.

In recent years, the topic of emotion reading has increasingly received attention from researchers in Cognitive Science and Artificial Intelligence. To study this phenomenon, in this paper an adaptive agent model is presented with capabilities to interpret another agent’s emotions. The presented agent model is based on recent advances in neurological context. First a non-adaptive agent model for emotion reading is described involving (preparatory) mirroring body states of the other agent. Here emotion reading is modelled taking into account the Simulation Theory perspective as known from the literature, involving the own body states and emotions in reading somebody else’s emotions. This models an agent that first develops the same feeling, and after feeling the emotion imputes it to the other agent. Next the agent model is extended to an adaptive model based on a Hebbian learning principle to develop a direct connection between a sensed stimulus concerning another agent’s body state (e.g., face expression) and the emotion recognition state. In this adaptive agent model the emotion is imputed to the other agent before it is actually felt. The agent model has been designed based on principles of neural modelling, and as such has a close relation to a neurological realisation.

The simulation of a large number of people’s evacuation behaviors is assumed to support the decision of rescue operations or prompt planning for disaster mitigation. Simulations of agents at wide areas with fine resolutions require a lot of computational resources and computer powers. We propose a hybrid traffic simulation combining network and area models to simulate agents’ behaviors. This paper presents that a hybrid traffic simulator had same results that have used all free space models. This indicates that our system can simulate behaviors a huge number of agents at wide areas with high resolutions by reasonable computational resources.

Agent-based simulation has demonstrated its usefulness for the modelling of complex systems. However, the simulation widely depends on the agent behaviour designing. In order to facilitate the definition of such behaviour, we propose an approach based on a participatory method: a domain expert directly enters his knowledge about entities in a specific environment. In this paper, we propose to formalise the agent behaviour by using a combination of production rules and of a multi-criteria decision making method. An experiment, carried out in the domain of ecological simulation, is presented. This first experiment shows promising results for our approach.

Network Topology has been shown as a key factor on influencing system performance even under the same coordination algorithm. Although many distributed algorithm designs have been proved to be feasible to make up some functions in the large scale distributed systems as claimed, for example, recovering the network from link or node failures, they may significantly change the network topology which has never been tested. Therefore, their influences on the overall system performance are unknown. In this paper, we have made an initial effort to find how a standard network recovery policy, such as MPLS algorithm, may change the network in terms of network congestions and shifts of network topologies. Those interesting discoveries are helpful to predict their influences on system performance and in turn to be useful for new algorithm design.

This paper addresses the problem of exploring how organizational structures may evolve over time using the information from the agents’ trust models. We present a mechanism based on clustering techniques capable of detecting behavioural patterns in organizational multi-agent systems, thereby identifying new roles that dynamically extend the role taxonomy. We present experimental results showing that this extension leads to an improvement of the agents’ decision making processes when compared to static organizational structures.

Human performance can seriously degrade under demanding tasks. To improve performance, agents can reason about the current state of the human, and give the most appropriate and effective support. To enable this, the agent needs a model of a specific person’s functional state and performance, which should be valid, as the agent might otherwise give inappropriate advice and even worsen performance. This paper concerns the adaptation of the parameters of the existing functional state model to the individual and validation of the resulting model. First, human experiments have been conducted, whereby measurements related to the model have been performed. Next, this data has been used to obtain appropriate parameter settings for the model, describing the specific subject. Finally, the model, with the tailored parameter settings, has been used to predict human behavior to investigate predictive capabilities of the model. The results have been analyzed using formal verification.

We present a new trading scheme in e-commerce in which end-users behave both buyers and sellers. We define hybrid traders as new users, analyze their trading models, and develop a trading mechanism. In our trading scheme, hybrid traders forge coalition formations to purchase items, since hybrid traders do not have enough money purchase extensive items. We create an incentive for coalitions using a side payment policy. We propose a side payment value decision mechanism based on the coalition’s contribution. Also, in multiple-item trading, we discuss the strategy analyses.

Simulation-based studies have attracted the attention of researchers of social science like politics and economics, and many findings are given in those domains through simulations. In spite of successful results, it has long been pointed out that there are relatively high bars for those researchers to newly adopt these approaches. We consider that one of them is the large cost related to programming, which is not only learning it but also understanding other researchers’ work represented both in their papers and in their codes. Another is the cost for introducing and managing computer systems required to perform efficient researches. We propose a system, GPGCloud, which helps users to share their simulation models with others, to cope with the programming related cost. It also provides an easy access to a large computation environment that allows them to simultaneously perform a number of simulation runs based on their models, which reduces the cost on managing large computers. We implemented a prototype of GPGCloud, which consists of several parts built on top of well-known open-source softwares. This paper describes the design and the implementation of the system.

Reputation mechanisms have been developed during last few years as valid methods to allow agents to better select partners in organisational environments. In most of works presented in the literature, reputation is summarised as a value, typically a number, that represents an opinion sent by an agent to another about a certain third party. In this work, we put forward a novel concept of

reputation-based agreement

in order to support the reputation definition, as well as, some desirable properties about it. We define a reputation service that collects opinions from agents, so creating

agreements

over

situations

. This service will also be in charge of presenting the information by using different

informative mechanisms

. Finally, a case study is presented in order to exemplify our work.

To develop a multilingual application based on the Language Grid, developers need to fill semantical gaps that exist among multilingual resources and software that are developed for different objects by various organizations. Authors have proposed a metadata-based approach, based on which the

Language Grid Facilitator

(LGF) is developed as a directory service that provides computational supports, and is aimed at helping developers who implement parallel text applications.

To cope with the increase of computational costs resulting from logical or semantical calculation performed in the LGF, we divided queries into two parts. One of them encodes static knowledge about the semantical gaps between resources and applications, and can be cached and indexed for better performance. A prototype was implemented, and experimental results have shown that our approach successfully reduced the computational cost while helping users to specify fine-grained conditions needed in the development of parallel text applications.

Here we introduce a novel algorithm for continual optimisation of dynamic distributed constraint optimisation problems. By using techniques derived from argumentation for communication the algorithm does not need to use an ordering over the variables. The lack of a hierarchy allows the algorithm to efficiently solve dynamic problems, as well as be completely asynchronous, fault tolerant and anytime. However it prevents an ordered search, making the algorithm incomplete.

In this paper, we analyzed the influence of time required to begin coping behaviors of managers in chemical terrorism. In order to calculate the damage of chemical attacks in a major rail station, our network model-based pedestrian simulator was applied with hazard prediction systems of indoor gas diffusion. Our analysis was used for enlightening the managers of the rail station in a tabletop exercise held by Kitakyushu City Fire and Disaster Management Department.