Skip to main content
main-content

Über dieses Buch

This book constitutes the proceedings of the 14th International Conference on Principles and Practice in Multi-Agent Systems, PRIMA 2011, held in Wollongong, Australia, in November 2011.

The 39 papers presented together with 3 invited talks were carefully reviewed and selected from numerous submissions. They focus on practical aspects of multiagent systems and are organised in topical sections on coalitions and teamwork, learning, mechanisms and voting, modeling and simulation, negotiation and coalitions, optimization, sustainability, agent societies and frameworks, argumentation, and applications.

Inhaltsverzeichnis

Frontmatter

Invited Talks

Game Theory for Security: Lessons Learned from Deployed Applications

Security at major locations of economic or political importance or transportation or other infrastructure is a key concern around the world, particularly given the threat of terrorism. Limited security resources prevent full security coverage at all times; instead, these limited resources must be deployed intelligently taking into account differences in priorities of targets requiring security coverage, the responses of the adversaries to the security posture and potential uncertainty over the types of adversaries faced. Game theory is well-suited to adversarial reasoning for security resource allocation and scheduling problems. Casting the problem as a Bayesian Stackelberg game, we have developed new algorithms for efficiently solving such games to provide randomized patrolling or inspection strategies: we can thus avoid predictability and address scale-up in these security scheduling problems, addressing key weaknesses of human scheduling. Our algorithms are now deployed in multiple applications. ARMOR, our first game theoretic application, has been deployed at the Los Angeles International Airport (LAX) since 2007 to randomize checkpoints on the roadways entering the airport and canine patrol routes within the airport terminals. IRIS, our second application, is a game-theoretic scheduler for randomized deployment of the Federal Air Marshals (FAMS) requiring significant scale-up in underlying algorithms; IRIS has been in use since 2009. Similarly, a new set of algorithms are deployed in Boston for a system called PROTECT for randomizing US coast guard patrolling; PROTECT is intended to be deployed at more locations in the future, and GUARDS is under evaluation for national deployment by the Transportation Security Administration (TSA). These applications are leading to real-world use-inspired research in scaling up to large-scale problems, handling significant adversarial uncertainty, dealing with bounded rationality of human adversaries, and other fundamental challenges. This talk will outline our algorithms, key research results and lessons learned from these applications.

Milind Tambe

Tools for a Robust, Sustainable Agent Community

We believe that intelligent information agents will represent their users interest in electronic marketplaces and other forums to trade, exchange, share, identify, and locate goods and services. Such information worlds will present unforeseen opportunities as well as challenges that can be best addressed by robust, self-sustaining agent communities. An agent community is a stable, adaptive group of self-interested agents that share common resources and must coordinate their efforts to effectively develop, utilize and nurture group resources and organization. More specifically, agents will need mechanisms to benefit from complementary expertise in the group, pool together resources to meet new demands and exploit transient opportunities, negotiate fair settlements, develop norms to facilitate coordination, exchange help and transfer knowledge between peers, secure the community against intruders, and learn to collaborate effectively. In this talk, I will summarize some of our research results on trust-based computing, negotiation, and learning that will enable intelligent agents to develop and sustain robust, adaptive, and successful agent communities.

Sandip Sen

From Notions to Models and Back Again, Again

A typical thread in research relies on the maturing of ideas through an iterative process of construction, testing and refinement. In this talk I will trace some such trajectories of ideas by illustration from some of my own and others’ experiences in agent-based modelling. I draw inspiration from previous commentaries, including from those who generated the mottos: “No Notation without Denotation” [1]; “No Notation without Exploitation” [2]; and “No experimentation without explanation” [3].

Liz Sonenberg

Full Papers

Coalitions and Teamwork

A Compact Representation Scheme of Coalitional Games Based on Multi-Terminal Zero-Suppressed Binary Decision Diagrams

Coalitional games, including Coalition Structure Generation (CSG), have been attracting considerable attention from the AI research community. Traditionally, the input of a coalitional game is a black-box function called a characteristic function. Previous studies have found that many problems in coalitional games tend to be computationally intractable in this black-box function representation. Recently, several concise representation schemes for a characteristic function have been proposed. Among them, a synergy coalition group (SCG) has several good characteristics, but its representation size tends to be large compared to other representation schemes.

We propose a new concise representation scheme for a characteristic function based on a Zero-suppressed Binary Decision Diagram (ZDD) and a SCG. We show our scheme (i) is fully expressive, (ii) can be exponentially more concise than the SCG representation, (iii) can solve core-related problems in polynomial time in the number of nodes, and (iv) can solve a CSG problem reasonably well by utilizing a MIP formulation. A Binary Decision Diagram (BDD) has been used as unified infrastructure for representing/manipulating discrete structures in such various domains in AI as data mining and knowledge discovery. Adapting this common infrastructure brings up the opportunity of utilizing abundant BDD resources and cross-fertilization with these fields.

Yuko Sakurai, Suguru Ueda, Atsushi Iwasaki, Shin-Ichi Minato, Makoto Yokoo

Environment Characterization for Non-recontaminating Frontier-Based Robotic Exploration

This paper addresses the problem of obtaining a concise description of a physical environment for robotic exploration. We aim to determine the number of robots required to clear an environment using non-recontaminating exploration. We introduce the medial axis as a configuration space and derive a mathematical representation of a continuous environment that captures its underlying topology and geometry. We show that this representation provides a concise description of arbitrary environments, and that reasoning about points in this representation is equivalent to reasoning about robots in physical space. We leverage this to derive a lower bound on the number of required pursuers. We provide a transformation from this continuous representation into a symbolic representation. Finally, we present a generalized pursuit-evasion algorithm. Given an environment we can compute how many pursuers we need, and generate an optimal pursuit strategy that will guarantee the evaders are detected with the minimum number of pursuers.

Mikhail Volkov, Alejandro Cornejo, Nancy Lynch, Daniela Rus

Learning

Aspects of Active Norm Learning and the Effect of Lying on Norm Emergence in Agent Societies

Norms have facilitated smoother functioning in human societies. In the field of normative multi-agent systems researchers are interested in investigating how the concept of social norms can be used to facilitate social order in electronic agent societies. In this context, the area of norm emergence has attracted a lot of interest among researchers. The objectives of this paper are two-fold. First, we discuss the norm learning approaches in agent societies and discuss the three aspects of active norm learning (experiential, observational and communication-based learning) in agent societies. Using an example we demonstrate the usefulness of combining these three aspects of norms learning. Second, we investigate the effect of the presence of liars in an agent society on norm emergence. Agents that lie distort truth when they are asked about the norm in an agent society. We show that lying has deleterious effect on norm emergence. In particular, using simulations we identify conditions under which the norms that have emerged in a society can be sustained in the presence of liars.

Bastin Tony Roy Savarimuthu, Rexy Arulanandam, Maryam Purvis

Mechanisms and Voting

Strategy-Proof Mechanisms for Interdependent Task Allocation with Private Durations

Classical mechanism design assumes that an agent’s value of any determined outcome depends only on its private information. However in many situations, an agent’s value of an outcome depends on the private information of other agents in addition to its private information. In such settings where agents have interdependent valuations, strategy-proof mechanisms have not been proposed yet, and when these mechanisms are possible is still an open research question. Toward addressing this question, we consider the interdependent task allocation (ITA) problem, where a set of tasks with predefined dependencies is to be assigned to self-interested agents based on what they report about their privately known capabilities and costs. We consider here the possibility that tasks may fail during their executions, which imposes interdependencies between the agents’ valuations. In this study, we design mechanisms and prove their strategy-proofness along with other properties for a class of ITA settings where an agent’s privately known costs are modeled as privately known durations.

Ayman Ghoneim

Costly Voting with Sequential Participation

This paper examines the property of the

m

votes to win mechanism. Voting is an effective way to make a collective decision but voting behaviors, e.g., monitoring the voting process, may incur a cost, that is, voting is often costly. In this case, compulsory voting incurs a larger cost. Random decision making can reduce the cost for voting but is skeptical in the quality of decision making. That is, we face the problem of how to balance the quality of collective decision making with the reduction of the cost for voting. To solve this problem, this paper focuses on the

m

votes to win mechanism, in which voters sequentially vote and if an alternative receives

m

votes, the voting process immediately terminates and the alternative received

m

votes wins. The similar voting mechanism is actually used in the Apache projects. However, the property of the

m

votes to win mechanism has not sufficiently studied. The questions include how to find a desirable value of

m

and what situation this mechanism is superior to other mechanisms. To answer these question, we create the discussion model where two alternatives is included, and analyze what voting strategy is rational. Based on the analysis, we examine what factors affects the social surplus, i.e., to what extent the quality of collective decision making and the reduction of the cost for voting are well balanced, and clarify whether the

m

votes to win mechanism is superior to the compulsory voting or the random decision making in terms of social surplus.

Ryuya Kagifuku, Shigeo Matsubara

Modelling and Simulation

An Agent-Based Model for Integrated Contagion and Regulation of Negative Mood

Through social interaction, the mood of a person can affect the mood of others. The speed and intensity of such mood contagion can differ, depending on the persons and the type and intensity of their interactions. Especially in close relationships the negative mood of a depressed person can have a serious impact on the moods of the ones close to him or her. For short time durations, contagion may be the main factor determining the mood of a person; however, for longer time durations individuals also apply regulation mechanisms to compensate for too strong deviations of their mood. Computational contagion models usually do not take into account such regulation. This paper introduces an agent-based model that simulates the spread of negative mood amongst a group of agents in a social network, but at the same time integrates elements from Gross’ emotion regulation theory, as the individuals’ efforts to avoid a negative mood. Simulation experiments under different group settings pointed out that the model is able to produce realistic results, that explain negative mood contagion and emotion regulation behaviours posed in the literature.

Azizi Ab Aziz, Jan Treur, C. Natalie van der Wal

A Framework for Agent-Based Modeling of Intelligent Goods

The purpose of this paper is to present a framework for intelligent goods and illustrate how it can be applied when modeling intelligent goods as agents. It includes a specification of different levels of capability connected to the goods, which is based on a number of capability dimensions. Additionally, three specific intelligent goods services related to transport are presented. We show how these services can be modeled as agents and how they relate to the intelligent goods framework. A discussion of different physical locations of service information and processing is also included.

Åse Jevinger, Paul Davidsson, Jan A. Persson

Multi-Agent Systems for Biomedical Simulation: Modeling Vascularization of Porous Scaffolds

An interesting application of multi-agent systems (MAS) is in modeling systems that can be represented by independent entities interacting together, the so-called agent-based modeling (ABM). In this paper MAS paradigm is used as a promising technique for representing complex biomedical systems. A brief survey of some ABM of biomedical systems is presented, followed by the description of a multi-layered agent-based framework developed in our own labs to model the process of sprouting angiogenesis (blood vessel formation) within polymeric porous scaffolds used for regenerative medicine. The ABM structure developed and challenges in modeling systems with a large number of rapidly increasing interacting agents are discussed. 2D and 3D case studies are presented to investigate the impact of scaffold pore structure on vessel growth. MAS provides a valuable tool for studying highly complex biological and biomedical systems, and for investigating ways of intervening in such systems.

Hamidreza Mehdizadeh, Arsun Artel, Eric M. Brey, Ali Cinar

Group Abstraction for Large-Scale Agent-Based Social Diffusion Models with Unaffected Agents

In this paper an approach is proposed to handle complex dynamics of large-scale multi-agents systems modelling social diffusion processes. A particular type of systems is considered, in which some agents (e.g., leaders) are not open to influence by the other agents. Based on local properties characterising the dynamics of individual agents and their interactions, groups and properties of the dynamics of these groups are identified. To determine such dynamic group properties two abstraction methods are proposed: determining group equilibrium states and approximation of group processes by weighted averaging of the interactions within the group. This enables simulation of the group dynamics at a more abstract level by considering groups as single entities substituting a large number of interacting agents. In this way the scalability of large-scale simulation can be improved significantly. Computational properties of the developed approach are addressed in the paper. The approach is illustrated for a collective decision making model with different types of topology, which may occur in social systems.

Alexei Sharpanskykh, Jan Treur

Negotiation

Towards a Quantitative Concession-Based Classification Method of Negotiation Strategies

In order to successfully reach an agreement in a negotiation, both parties rely on each other to make concessions. The willingness to concede also depends in large part on the opponent. A concession by the opponent may be reciprocated, but the negotiation process may also be frustrated if the opponent does not concede at all.

This process of concession making is a central theme in many of the classic and current automated negotiation strategies. In this paper, we present a quantitative classification method of negotiation strategies that measures the willingness of an agent to concede against different types of opponents. The method is then applied to classify some well-known negotiating strategies, including the agents of ANAC 2010. It is shown that the technique makes it easy to identify the main characteristics of negotiation agents, and can be used to group negotiation strategies into categories with common negotiation characteristics. We also observe, among other things, that different kinds of opponents call for a different approach in making concessions.

Tim Baarslag, Koen Hindriks, Catholijn Jonker

Consensus Policy Based Multi-agent Negotiation

Multiagent negotiation may be understood as a consensus based group decision-making which ideally should seek the agreement of all the participants. However, there exist situations where an unanimous agreement is not possible or simply the rules imposed by the system do not seek such unanimous agreement. In this paper we propose to use a consensus policy based mediation framework (CPMF) to perform multiagent negotiations. This proposal fills a gap in the literature where protocols are in most cases indirectly biased to search for a quorum. The mechanisms proposed to perform the exploration of the negotiation space are derived from the Generalized Pattern Search non-linear optimization technique (GPS). The mediation mechanisms are guided by the aggregation of the agent preferences on the set of alternatives the mediator proposes in each negotiation round. Considerable interest is focused on the implementation of the mediation rules where we allow for a linguistic description of the type of agreements needed. We show empirically that CPMF efficiently manages negotiations following predefined consensus policies and solves situations where unanimous agreements are not viable.

Enrique de la Hoz, Miguel A. Lopez-Carmona, Mark Klein, Ivan Marsa-Maestre

Optimization

Distributed Lagrangian Relaxation Protocol for the Over-constrained Generalized Mutual Assignment Problem

The Generalized Mutual Assignment Problem

(GMAP) is a distributed combinatorial optimization problem in which, with no centralized control, multiple agents search for an optimal assignment of goods that satisfies their individual knapsack constraints. Previously, in the GMAP protocol, problem instances were assumed to be feasible, meaning that the total capacities of the agents were large enough to assign the goods. However, this assumption may not be realistic in some situations. In this paper, we present two methods for dealing with such “over-constrained” GMAP instances. First, we introduce a

disposal agent

who has an unlimited capacity and is in charge of the unassigned goods. With this method, we can use any off-the-shelf GMAP protocol since the disposal agent can make the instances feasible. Second, we formulate the GMAP instances as an Integer Programming (IP) problem, in which the assignment constraints are described with inequalities. With this method, we need to devise a new protocol for such a formulation. We experimentally compared these two methods on the variants of

Generalized Assignment Problem

(GAP) benchmark instances. Our results indicate that the first method finds a solution faster for fewer over-constrained instances, and the second finds a better solution faster for more over-constrained instances.

Kenta Hanada, Katsutoshi Hirayama

The Effect of Congestion Frequency and Saturation on Coordinated Traffic Routing

Traffic congestion is a widespread epidemic that continually wreaks havoc in urban areas. Traffic jams, car wrecks, construction delays, and other causes of congestion, can turn even the biggest highways into a parking lot. Several congestion mitigation strategies are being studied, many focusing on micro-simulation of traffic to determine how modifying road structures will affect the flow of traffic and the networking perspective of vehicle-to-vehicle communication. Vehicle routing on a network of roads and intersections can be modeled as a distributed constraint optimization problem and solved using a range of centralized to decentralized techniques. In this paper, we present a constraint optimization model of a traffic routing problem. We produce congestion data using a sinusoidal wave pattern and vary its amplitude (saturation) and frequency (vehicle waves through a given intersection). Through empirical evaluation, we show how a centralized and decentralized solution each react to unknown congestion information that occurs after the initial route planning period.

Melanie Smith, Roger Mailler

Sustainability

Coordination, Conventions and the Self-organisation of Sustainable Institutions

Applications where autonomous and heterogeneous agents form opportunistic alliances, which require them to share collective resources to achieve individual objectives, are increasingly common. We model such applications in terms of self-governing institutions for shared resource management. Socio-economic principles for enduring institutions are formalised in a logical framework for dynamic specification of norm-governed systems. The framework is implemented in an experimental testbed to investigate the interplay of coordination in a social dilemma with mutable conventions of an institution. Experimental results show that the presence of conventions enables the norm-governed system to approximate the performance of a theoretically ideal system. We conclude that this approach to self-organisation can provide the foundations for implementing sustainable electronic institutions.

Jeremy Pitt, Julia Schaumeier, Alexander Artikis

An Agent-Based Extensible Climate Control System for Sustainable Greenhouse Production

The slow adoption pace of new control strategies for sustainable greenhouse climate control by industrial growers, is mainly due to the complexity of identifying and resolving potentially conflicting climate control requirements. In this paper, we present a multi-agent-based climate control system that allows new control strategies to be adopted without any need to identify or resolve conflicts beforehand. This is achieved by representing the climate control requirements as separate agents. Identifying and solving conflicts then becomes a negotiation problem among agents sharing the same controlled environment. Negotiation is done using a novel multi-objective negotiation protocol that uses a generic algorithm to find an optimized solution within the search space. The multi-agent-based control system has been empirically evaluated in an ornamental floriculture research facility in Denmark. The evaluation showed that it is realistic to implement the climate control requirements as individual agents, thereby opening greenhouse climate control systems for integration of independently produced control strategies.

Jan Corfixen Sørensen, Bo Nørregaard Jørgensen, Mark Klein, Yves Demazeau

Applications

ACTraversal: Ranking Crowdsourced Commonsense Assertions and Certifications

Building commonsense knowledge bases is a challenging undertaking. While we have witnessed the successful collection of large amounts of commonsense knowledge by either automatic text mining or

games with a purpose

(GWAP), such data are of limited precision. Verifying data is typically done with repetition, which works better for very large data sets. Our research proposes a novel approach to data verification by coupling multiple data collection methods. This paper presents

ACTraversal

, a graph traversal algorithm for ranking data collected from GWAP and text mining. Experiments on aggregating data from two GWAPs, i.e. Virtual Pets and Top10, with two text mining tools, i.e. SEAL and Google Distance, showed significant improvements.

Tao-Hsuan Chang, Yen-Ling Kuo, Jane Yung-jen Hsu

Automated Adaptation of Strategic Guidance in Multiagent Coordination

We address multi-agent planning problems in dynamic environments motivated by assisting human teams in disaster emergency response. It is challenging because most goals are revealed during execution, where uncertainty in the duration and outcome of actions plays a significant role, and where unexpected events can cause large disruptions to existing plans. The key to our approach is giving human planners a rich strategy language to constrain the assignment of agents to goals and allow the system to instantiate the strategy during execution, tuning the assignment to the evolving execution state. Our approach outperformed an extensively-trained team coordinating with radios and a traditional command-center organization, and an agent-assisted team using a different approach.

Rajiv T. Maheswaran, Pedro Szekely, Romeo Sanchez

Early Innovation Papers

Agent Societies and Frameworks

Weaving a Fabric of Socially Aware Agents

The expansion of web-enabled social interaction has shed light on social aspects of intelligence that have not been typically studied within the AI paradigm so far. In this context, our aim is to understand what constitutes intelligent social behaviour and to build computational systems that support it. We argue that social intelligence involves socially aware, autonomous individuals that agree on how to accomplish a common endeavour, and then enact such agreements. In particular, we provide a framework with the essential elements for such agreements to be achieved and executed by individuals that meet in an open environment. Such framework sets the foundations to build a computational infrastructure that enables socially aware autonomy.

Mark d’Inverno, Michael Luck, Pablo Noriega, Juan A. Rodriguez-Aguilar, Carles Sierra

Dynamic Ad Hoc Coordination of Distributed Tasks Using Micro-Agents

The notion of

μ

-agents to develop complex software applications has been under active research interest for some time. Through improved organisational modelling

μ

-agents provide stronger support for decomposition and abstraction in decentralized applications. With the advent of the mobile application platform Android – which exhibits strong analogies to multi-agent system principles – we strongly believe that

μ

-agent-based modelling has become an increasingly attractive alternative. It can combine decentralized application development with the wide-ranging set of sensors and communication channels to foster both context-sensitivity and flexibility of applications. By integrating Android with the

μ

-agent concept mobile applications can put stronger emphasis on coordination of task-oriented agent organisations. As an example how this can facilitate the development of distributed applications, we describe an application for the field of ”Unconferences” to dynamically schedule informal talks in an ad hoc manner. We model the central aspects of the application and show the advantages of our

μ

-agent-based approach. Finally, we contrast our approach to existing work in this field and suggest the consideration of

μ

-agents as an alternative to conventional object-oriented software development.

Christopher Frantz, Mariusz Nowostawski, Martin K. Purvis

Programming Dynamics of Multi-Agent Systems

Dynamics are one of the most important properties of multi-agent systems (MAS), which often operate in open environment and with dynamically changing requirements. This paper firstly gives a comprehensive view of the dynamics in MAS based on “where” and “what” aspects of change and discusses the software engineering issues of engineering such dynamics. To solve related issues, we propose an organization-based programming approach that provides programming abstraction and mechanisms to describe and manage dynamics of MAS. An organization-based language for programming dynamics (OBLPD) of MAS is defined. The syntax of OBLPD is defined and its semantics are informally explained with a case study.

Cuiyun Hu, Xinjun Mao, Huiping Zhou

Capability Modeling of Knowledge-Based Agents for Commonsense Knowledge Integration

Robust intelligent systems require commonsense knowledge. While significant progress has been made in building large commonsense knowledge bases, they are intrinsically incomplete. It is difficult to combine multiple knowledge bases due to their different choices of representation and inference mechanisms, thereby limiting users to one knowledge base and its reasonable methods for any specific task. This paper presents a multi-agent framework for commonsense knowledge integration, and proposes an approach to capability modeling of knowledge bases without a common ontology. The proposed capability model provides a general description of large heterogeneous knowledge bases, such that contents accessible by the knowledge-based agents may be matched up against specific requests. The concept correlation matrix of a knowledge base is transformed into a

k

-dimensional vector space using low-rank approximation for dimensionality reduction. Experiments are performed with the matchmaking mechanism for commonsense knowledge integration framework using the capability models of ConceptNet, WordNet, and Wikipedia. In the user study, the matchmaking results are compared with the ranked lists produced by online users to show that over 85% of them are accurate and have positive correlation with the user-produced ranked lists.

Yen-Ling Kuo, Jane Yung-jen Hsu

Producing Enactable Protocols in Artificial Agent Societies

This paper draws upon our previous work [7, 16] in which we proposed the organisation of services around the concept of

artificial agent societies

and presented a framework for representing roles and protocols using LTSs. The agent would apply for a role in the society, which would result in its participation in a number of protocols. We advocated the use of the

games-based metaphor

for describing the protocols and presented a framework for assessing the admission of the agent to the society on the basis of its

competence

. In this work we look at the subsequent question:

what information should the agent receive upon entry?

. We can not provide it with the full protocol because of security and overload issues. Therefore, we choose to only provide the actions pertinent to the protocols that the role the agent applied for participates in the society. We employ

branching bisimulation

for producing a protocol equivalent to the original one with all actions not involving the role translated into silent (

τ

) actions. However, this approach sometimes results in

non-enactable

protocols. In this case, we need to

repair

the protocol by adding the role in question as a recipient to certain protocol messages that were causing the problems. We present three different approaches for repairing protocols, depending on the number of messages from the original protocol they modify. The modified protocol is adopted as the final one and the agent is given the role automaton that is derived from the

branching bisimulation

process.

George K. Lekeas, Christos Kloukinas, Kostas Stathis

Argumentation

Argumentation Schemes for Collaborative Planning

We address the collaborative planning problem among agents where they have different objectives and norms. In this context, agreeing on the best course of action to adopt represents a significant challenge. Concurrent actions and causal plan-constraints may lead to conflicts of opinion on what to do. Moreover, individual norms can constrain agent behaviour. We propose an argumentation-based model for deliberative dialogues based on argumentation schemes. This model facilitates agreements about joint plans by enriching the quality of the dialogue through the exchange of relevant information about plan commitments and norms.

Alice Toniolo, Timothy J. Norman, Katia Sycara

Preference-Based Argumentation Handling Dynamic Preferences Built on Prioritized Logic Programming

To treat dynamic preferences correctly is crucially required in the fields of argumentation as well as nonmonotonic reasoning. To meet such requirements, first, we propose a

hierarchical

Prioritized Logic Program (or a

hierarchical

PLP, for short), which enhances the formalism of Sakama and Inoue’s PLP so that it can represent and reason about dynamic preferences. Second, using such a hierarchical PLP as the underlying language, the proposed method defines the preference-based argumentation framework (called the dynamic

PAF

) built from it. This enables us to argue and reason about dynamic preferences in argumentation. Finally we show the interesting relationship between semantics of a hierarchical PLP given by

preferred

answer sets and semantics of the dynamic

PAF

given by

${\cal P}$

-extensions.

Toshiko Wakaki

Learning

Adaption of Stepsize Parameter Using Newton’s Method

A method to optimize stepsize parameters in exponential moving average (EMA) based on Newton’s method to minimize square errors is proposed. The stepsize parameters used in reinforcement learning methods should be selected and adjusted carefully for dynamic and non-stationary environments. To find the suitable values for the stepsize parameters through learning, a framework to acquire higher-order derivatives of learning values by the stepsize parameters has been proposed. Based on this framework, the authors extend a method to determine the best stepsize using Newton’s method to minimize EMA of square error of learning. The method is confirmed by mathematical theories and by results of experiments.

Itsuki Noda

A Health Social Network Recommender System

People with chronic health conditions require support beyond normal health care systems. Social networking has shown great potential to provide the needed support. Because of the privacy and security issues of health information systems, it is often difficult to find patients who can support each other in the community. We propose a social-networking framework for patient care, in particular for parents of children with Autism Spectrum Disorders (ASD). In the framework, health service providers facilitate social links between parents using similarities of assessment reports without revealing sensitive information. A machine learning approach was developed to generate explanations of ASD assessments in order to assist clinicians in their assessment. The generated explanations are then used to measure similarities between assessments in order to recommend a community of related parents. For the first time, we report on the accuracy of social linking using an explanation-based similarity measure.

Insu Song, Denise Dillon, Tze Jui Goh, Min Sung

Modelling

Learning Belief Connections in a Model for Situation Awareness

Situational awareness is critical in many human tasks, especially in cases where humans have to make decisions fast and where the result of their decisions might affect their life. This paper addresses the problem of learning optimal values for the parameters of a situational awareness model. The model is a complex network with nodes connected by links with weights, which connect observations to simple beliefs, such as “there is a contact”, to complex belief, such as “the contact is hostile”, and to future beliefs, such as “it is possible the pilot is being targeted”. The model has been built and validated by human experts in the domain of F16 fighter pilots and is used to study human decision making. Given the complexity of the model, there is a need to learn appropriate weights for the connections, which, in turn, affect the activation levels of the beliefs. We propose the use of a genetic algorithm and of a sensitivity based approach to learn the weights in the model. Extensive experimental results are included.

Maria L. Gini, Mark Hoogendoorn, Rianne van Lambalgen

An Integrated Agent Model Addressing Situation Awareness and Functional State in Decision Making

In this paper, an integrated agent model is introduced addressing mutually interacting Situation Awareness and Functional State dynamics in decision making. This shows how a human’s functional state, more specific a human’s exhaustion and power, can influence a human’s situation awareness, and in turn the decision making. The model is illustrated by a number of simulation scenarios.

Mark Hoogendoorn, Rianne van Lambalgen, Jan Treur

Agent-Based Modelling for Understanding Sustainability

Our aim is to demonstrate how agent-based models can play an important role in understanding sustainability. Here, we describe how the agent-based motivation models support the description of desirable outcomes and help to develop relevant and shared high-level goals in particularly complex areas such as sustainability. We focus on sustainable behaviour in households, and how to provide guidance for people to behave in a more environmentally-friendly manner. Our example demonstrates that the agent-based models are able to focus on the right questions when making decisions between alternatives in this multifaceted domain. With agent-based models, we aim to enable people to make the right choices for their particular circumstances and values. The agent paradigm is uniquely suited to understanding and representing the relevant goals, quality goals, and individual activities.

Sonja Pedell, Leon Sterling

Modelling Joint Decision Making Processes Involving Emotion-Related Valuing and Empathic Understanding

In this paper a social agent model for joint decision making is presented addressing the role of mutually acknowledged empathic understanding in the decision making. The model is based on principles from recent neurological theories on mirror neurons, internal simulation, and emotion-related valuing. Emotion-related valuing of decision options and mutual contagion of intentions and emotions between agents are used as a basis for mutual empathic understanding and convergence of decisions and their associated emotions.

Jan Treur

Negotiation and Coalitions

Negowiki: A Set of Community Tools for the Consistent Comparison of Negotiation Approaches

There is a number of recent research lines addressing automated complex negotiations. Most of them focus on overcoming the problems imposed by the complexity of negotiation scenarios which are computationally intractable, be it by approximating these complex scenarios with simpler ones, or by developing heuristic mechanisms to explore more efficiently the solution space. The problem with these mechanisms is that their evaluation is usually restricted to very specific negotiation scenarios, which makes very difficult to compare different approaches, to re-use concepts from previous mechanisms to create new ones or to generalize mechanisms to other scenarios. This makes the different research lines in automated negotiation to progress in an isolated manner. A solution to this recurring problem might be to create a collection of negotiation scenarios which may be used to benchmark different negotiation approaches. This paper aims to fill this gap by providing a framework for the characterization and generation of negotiation scenarios intended to address this problem. The framework has been integrated in a website, called the

Negowiki

, which allows to share scenarios and experiment results with the negotiation community, facilitating in this way that researchers compare and share their advancements.

Ivan Marsa-Maestre, Mark Klein, Enrique de la Hoz, Miguel A. Lopez-Carmona

A Stochastic Negotiation Approach to Power Restoration Problems in a Smart Grid

In this paper we propose a negotiation protocol for multi-feeder agents who must resolve the conflicts in order to find an optimal allocation of power in order for the power restoration after a blackout in a power grid. Since the power distribution domain constraints and cost, the optimal power distribution criteria involves multi-objectives such as number of changes of switches, number of power zones restored, and etc. It is not usually easy to come up with an optimal solution in a scaled-up complicated power grid topology within a limited time. We implemented two stochastic decision functions PDF and ZDF for feeder agents to decide whether to accept a proposal of other feeder agent who requests a restoration zone and whether to request a candidate target zone to deliver the power respectively in the negotiation. We show that with ZDF the feeder agents can negotiate faster to come up with the optimal solution than without ZDF. Finally we also show that our negotiation is a real time algorithm and show the performance curve of negotiation in terms of the restoration rate.

Von-Wun Soo, Yen-Bo Peng

Coalition-Oriented Sensing in Wireless Sensor Networks

Wireless Sensor Networks are generally composed of a large number of nodes that monitor their surrounding area. The monitoring capacity of sensors gets altered by the changing conditions of the environment and the sensors’ internal state. Sensor coalitions, in which only the leader transmits information to a sink node, are a means to save resources when the conditions of the environment are similar around the sensors in the coalition. In this paper we analyse and formalise such sensor coalitions and propose an algorithm for coalition formation that allows the sensors to self-organise with the purpose of performing a good monitoring of the environment while maximising the life span of the sensor network as a whole. The algorithm uses the quality of the information fused at the coalition leader and the remaining energy of the sensors as the basic parameters to alter coalition membership and leadership.

M. del Carmen Delgado-Roman, Carles Sierra

Simulation

Reducing the Environmental Impact of New Construction Projects through General Purpose Building Design and Multi-agent Crowd Simulation

This paper presents a two stage process for using intelligent agent technology in designing space-efficient buildings in order to reduce their environmental impact. An environment editor for designing new construction projects is described, followed by a microscopic crowd simulation model that is used to test the operational efficiency of the designed building. Each member of the crowd is represented as an intelligent agent, which allows for more complex goal-directed behaviour, which in turn leads to more realistic crowd behaviour. Crowd simulations can be used to detect potential problem areas, as well as identify areas that may safely be made smaller.

Kieron Ekron, Jaco Bijker, Elize Ehlers

An Investigation of Emergent Collaboration under Uncertainty and Minimal Information in Energy Domains

We study the phenomenon of evolution of cooperation in the electricity domain, where self-interested agents representing distributed energy resources (DERs) strategize for maximizing payoff. From the system’s viewpoint cooperation represents a solution capable to cope with the increasing complexity, generated by the introduction of DERs to the grid. The problem domain is modelled from a multi-agent system high-level perspective. We report on experiments with this model, giving the underlying understanding for the emergent behavior, in order to determine if and under what conditions such a collaborative behavior would hold. Finally we suggest how insights from this model can inspire mechanisms to instill cooperation as the dominant strategy.

Radu-Casian Mihailescu, Matteo Vasirani, Sascha Ossowski

A Simulator Integration Platform for City Simulations

Multiagent-based simulations are regarded as an useful technology for analyzing complex social systems and have been applied to various problems. Tackling the problems of a city involves various levels of abstraction and various target domains. Different types of human behaviors are studied separately by specialists in their respective domains. We believe that we need to integrate simulators that offer different levels of abstraction and cover various target domains. This paper introduces the architecture of a simulator integration platform and demonstrates the capability of the platform in that domains of city traffic and city electricity.

Yuu Nakajima, Hiromitsu Hattori

Human Relationship Modeling in Agent-Based Crowd Evacuation Simulation

Crowd evacuation simulations are becoming a tool to analyze and assess the safety of occupants in buildings. Agent-based simulation provides a platform on which to compute individual and collective behaviors that occur in crowds. We propose a human behavior model in evacuation based on the Belief-Desire-Intention (BDI) model and Helbing’s agent behavior model. Human relationships affect the states of BDI at each simulation step, and altruism forces among agents are introduced in Helbing’s model to affect agents’ intentions in calculating agent movements. Two evacuation scenarios are examined so that the results match quantitatively and qualitatively with past disasters. The simulations reveal typical behaviors in a crowd evacuation; for example, family-minded human behaviors that lead to interactions in the crowd and other behaviors. The simulation indicates that due to the interaction it takes a longer time to evacuate from buildings in actual situations.

Masaru Okaya, Tomoichi Takahashi

Applications

A Web Service Recommendation System Based on Users’ Reputations

In recent years, as the Internet spreads, the use of the Web Service has increased, and it has diversified. The Web Service is registered with UDDI, and the user selects service there and can use it for the provider by making a demand. In future, if the Web Service comes to be used more widely, the number of Web Services will increase, and the number of registrations at the UDDI will also increase. The user examines the large number of available services, and needs to choose the service that best matches their purpose. Quality of Service (QoS) is used as an index when a user chooses a service. Many studies show that the scoring of QoS for service selection is important. Quality of Service is registered by the provider and is treated as an objective factor. However, subjective evaluation, the evaluation of the user after the service use, is also needed to choose the best service. In this study, we use a new element, evaluation, in addition to QoS for service selection. We have expanded the existing filtering technique to make a new way of recommending services. Our method incorporates subjective evaluation. With this model, we apply the technique of information filtering to the Web Service recommendation and make an agent. Also, we simulate it after having clarified the behavior and tested it. The results of testing show that the model provides high levels of precision.

Yu Furusawa, Yuta Sugiki, Reiko Hishiyama

Human-Centered Planning for Adaptive User Situation in Ambient Intelligence Environment

The new intelligence system named Ambient Intelligence (AmI) is now in the limelight. In this AmI environment, when multiple people in various situations cooperate mutually, they need to to be supported simultaneously and effectively. Furthermore, peoples’ context and activities change continuously, so it is also necessary to achieve dynamic correspondence between people. Therefore, a planning agent has been developed to enable context-aware service composition. The planning agent creates plans dynamically to support multiple people in different environments. In the experiment, the dish event scenario was adopted and the planning agent allotted multiple people ingredient collection tasks.

Makoto Sando, Reiko Hishiyama

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise