Search

A result of Bulow and Klemperer has suggested that auctions may be a better tool to obtain an efficient outcome than negotiation. For example, some auction mechanisms can be shown to be efficient and strategy-proof. However, they generally also require that the preferences of at least one side of the auction are publicly known. However, sometimes it is very costly, impossible, or undesirable to publicly announce such preferences. It thus is interesting to find methods that do not impose this constraint but still approximate the outcome of the auction. In this paper we show that a multi-round multi-party negotiation protocol may be used to this end if the negotiating agents are capable of learning opponent preferences. The latter condition can be met by current state of the art negotiation technology. We show that this protocol approximates the theoretical outcome predicted by a so-called Qualitative Vickrey auction mechanism (even) on a complex multi-issue domain. [ABSTRACT FROM AUTHOR]

No intelligent decision support system functions even remotely without knowing the preferences of the user. A major problem is that the way average users think about and formulate their preferences does not match the utility-based quantitative frameworks currently used in decision support systems. For the average user qualitative models are a better fit. This paper presents an argumentation-based framework for the modelling of, and automated reasoning about multi-issue preferences of a qualitative nature. The framework presents preferences according to the lexicographic ordering that is well-understood by humans. The main contribution of the paper is that it shows how to reason about preferences when only incomplete information is available. An adequate strategy is proposed that allows reasoning with incomplete information and it is shown how to incorporate this strategy into the argumentation-based framework for modelling preferences. [ABSTRACT FROM AUTHOR]

In this paper we focus on explaining to humans the behavior of autonomous agents, i.e., explainable agents. Explainable agents are useful for many reasons including scenario-based training (e.g. disaster training), tutor and pedagogical systems, agent development and debugging, gaming, and interactive storytelling. As the aim is to generate for humans plausible and insightful explanations, user evaluation of different explanations is essential. In this paper we test the hypothesis that different explanation types are needed to explain different types of actions. We present three different, generically applicable, algorithms that automatically generate different types of explanations for actions of BDI-based agents. Quantitative analysis of a user experiment (n=30), in which users rated the usefulness and naturalness of each explanation type for different agent actions, supports our hypothesis. In addition, we present feedback from the users about how they would explain the actions themselves. Finally, we hypothesize guidelines relevant for the development of explainable BDI agents. [ABSTRACT FROM AUTHOR]

Model checking agent programs is a challenge and it is still a question which approaches can suitably be applied to effectively model check such programs. We present a new approach to explicit-state, on-the-fly model checking for agent programs. In this approach we use the agent program interpreter for generating the state space. A model checker is built on top of this interpreter by implementing efficient transformations of temporal properties to Büchi automata and an efficient bookkeeping mechanism that maintains track of states that have been visited. The proposed approach is generic and can be applied to different agent programming frameworks. We evaluate this approach to model checking by comparing it empirically with an approach based on the Maude model checker, and one based on the Agent Infrastructure Layer (AIL) intermediate language in combination with JPF. It turns out that although our approach does not use state-space reduction techniques, it shows significantly improved performance over these approaches. To the best of our knowledge, no such comparisons of approaches to model checking agent programs have been done before. [ABSTRACT FROM AUTHOR]

Communication is key in a multi-agent system for agents to exchange information and coordinate their activities. In the area of agent programming, the challenge is to introduce communication primitives that are useful to a programmer of agent programs as well as semantically well-defined. Moreover, for agents that derive their choice of action from their beliefs and goals it is natural to introduce primitives that support communication related to both of these attitudes. We introduce a communication approach for multi-agent systems based on mood operators instead of the usual speech act labels and a semantics based on the idea that a message can be used to (re)construct a mental model of the sender. An operational semantics is provided that specifies the precise meaning of the primitives. Finally, to facilitate coordination in multi-agent systems, we introduce the concept of a conversation to synchronize actions and communication among agents. Conversations provide a limited resource at the multi-agent level, and provide a natural approach for multi-agent systems to coordinate agent activities. [ABSTRACT FROM AUTHOR]

In automated negotiation, information gained about an opponent΄s preference profile by means of learning techniques may significantly improve an agent΄s negotiation performance. It therefore is useful to gain a better understanding of how various negotiation factors influence the quality of learning. The quality of learning techniques in negotiation are typically assessed indirectly by means of comparing the utility levels of agreed outcomes and other more global negotiation parameters. An evaluation of learning based on such general criteria, however, does not provide any insight into the influence of various aspects of negotiation on the quality of the learned model itself. The quality may depend on such aspects as the domain of negotiation, the structure of the preference profiles, the negotiation strategies used by the parties, and others. To gain a better understanding of the performance of proposed learning techniques in the context of negotiation and to be able to assess the potential to improve the performance of such techniques a more systematic assessment method is needed. In this paper we propose such a systematic method to analyse the quality of the information gained about opponent preferences by learning in single-instance negotiations. The method includes measures to assess the quality of a learned preference profile and proposes an experimental setup to analyse the influence of various negotiation aspects on the quality of learning. We apply the method to a Bayesian learning approach for learning an opponent΄s preference profile and discuss our findings. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

It has been observed that there are interesting relations between planning and agent programming. This is not surprising as agent programming was partially motivated by the lack of planners that are able to operate in dynamic, complex environments. Vice versa it has also been observed, however, that agent programming languages typically lack planning capabilities. We show in this paper that the agent programming language Goal is not only a programming language but can actually be used as a planning formalism as well. This opens up many possibilities for various approaches to mix execution and planning in agent-oriented programming. Moreover, by using the recently introduced temporalGoal we are able to include not only the stratified axioms and ADL that are part of PDDL but also plan constraints. [ABSTRACT FROM AUTHOR]

We present an abstract framework for verifying communicative actions for rational agent programming languages. Firstly, a multi-agent verification logic based on the computational semantics is introduced; and subsequently, this multi-agent logic is embedded into a more expressive modal logic over a standard run-based semantics. We formally relate both logics, prove expressivity results, and argue why it is useful to have a (more expressive) standard modal logic and semantics at hand. [ABSTRACT FROM AUTHOR]

Rational agents programmed in agent programming lan- guages derive their choice of action from their beliefs and goals. One of the main benefits of such programming languages is that they facilitate a high-level and conceptually elegant specification of agent behaviour. Qualitative concepts alone, however, are not sufficient to specify that this behaviour is also nearly optimal, a quality typically also associated with rational agents. Optimality in this context refers to the costs and rewards associated with action execution. It thus would be useful to extend agent programming languages with primitives that allow the specification of near-optimal behaviour. The idea is that quantitative heuristics added to an agent program prune some of the options generated by the qualitative action selection mechanism. In this paper, we explore the expressivity needed to specify such behaviour in the Blocks World domain. The programming constructs that we introduce allow for a high-level specification of such heuristics due to the fact that these can be defined by (re)using the qualitative notions of the basic agent programming language again. We illustrate the use of these constructs by extending a Goal Blocks World agent with various strategies to optimize its behaviour. [ABSTRACT FROM AUTHOR]

The core capability of a rational agent is to choose its next action in a rational fashion, a capability that can be put to good use by a designer to satisfy the design objectives of an agent system. In agent programming languages for rational agents, such choices are derived from the agent΄s beliefs and goals. At any one time, an agent can typically choose from multiple actions, which may all lead to goal achievement. Existing approaches usually select one of those actions non-deterministically. In this paper, we propose the use of goals as hard constraints and qualitative preferences as soft constraints for choosing a most preferred action among the available ones. We use temporal logic for the representation of various kinds of goals and preferences, leading to a uniform framework for integrating goals and preferences into agent programming. [ABSTRACT FROM AUTHOR]

In most agent programming languages in practice a programmer is committed to the use of a single knowledge representation technology. In this paper we argue this is not necessarily so. It is shown that rational agent programming languages allow for the combination of various such technologies. Specific issues that have to be addressed to realize such integration for rational agents that derive their choice of action from their beliefs and goals are discussed. Two techniques to deal with these issues which enable the integration of multiple knowledge representation techniques are presented: a meaning-preserving translation approach that maps one representation to another, and an approach based on so-called bridge rules which add additional inference power to a system combining multiple knowledge representation technologies. [ABSTRACT FROM AUTHOR]

Organizational notions such as roles, norms (e.g., obligations and permissions), and services are increasingly viewed as natural concepts to manage the complexity of software development. In particular in the context of multi-agent systems, agents are expected to be organization-aware, i.e., to understand and reason about the structure, work processes, and norms of the agent organization in which they operate. In this paper, we analyze which kinds of reasoning an agent should be able to do to function in an organization. We categorize these kinds of reasoning with respect to several dimensions, and distinguish three general approaches on how these might be integrated in existing agent programming languages. Through this, we provide a research agenda on what needs to be addressed when developing techniques for programming organization-aware agents. [ABSTRACT FROM AUTHOR]

A rational agent derives its choice of action from its beliefs and goals. Goals can be distinguished into achievement goals and maintenance goals. The aim of this paper is to define a mechanism which ensures the satisfaction of maintenance goals. We argue that such a mechanism requires the agent to look ahead, in order to make sure that the execution of actions does not lead to a violation of a maintenance goal. That is, maintenance goals may constrain the agent in choosing its actions. We propose a formal semantics of maintenance goals based on the notion of lookahead, and analyze the semantics by proving some properties. Additionally, we discuss the issue of achievement goal revision, in case the maintenance goals are so restrictive that all courses of action for satisfying achievement goals will lead to a violation of maintenance goals. [ABSTRACT FROM AUTHOR]

This paper presents the design of an open architecture for heterogeneous negotiating agents. Both the system level architecture as well as the architecture for negotiating agents are provided. The main contribution of this paper is that it derives a precisely specified interface from these architectures that facilitates an easy integration of heterogeneous agents into the overall negotiation framework. The interface is defined as a set of adapters that allows for various levels of integration of agents into the system architecture. The functionality provided by the system architecture depends on the number of adapters that are implemented and used to connect an agent to this architecture, ranging from functionality to conduct a bilateral negotiation to functionality for computing agent internal performance measures such as the quality of an opponent model. The architecture is used as the basis of a competitive testbed which allows us to study various negotiating agents. The design yields a flexible negotiation framework that facilitates negotiating different domains potentially using different protocols whereas no details of the internal negotiating agent structure are enforced. An application of the framework is illustrated by integrating two agents from the literature. [ABSTRACT FROM AUTHOR]

A variety of agent-oriented programming languages based on concepts such as beliefs and goals has been proposed in the literature. Even though most of these languages now come with interpreters implemented in e.g. Java and can be used to write software agents, there is little work reporting how to implement such languages or to identify a core instruction set that would facilitate such implementation. In this paper we introduce a compiler for the language GOAL into the framework of Jazzyk Behavioural State Machines. The result is a translation of key agent concepts such as beliefs and goals into Jazzyk which lacks these notions, thus providing some evidence that it may provide a sufficient instruction set for implementing agent programs. Moreover, arguably, the implementation strategy used can be applied also to other agent programming languages. [ABSTRACT FROM AUTHOR]

Modular programming has the usual benefits associated with structured programming, information hiding and reusability, but also has additional benefits to offer when applied in agent programming. We argue that modules can be viewed as structures similar to that of policy-based intentions [2]. Modules perceived in this way are components within an agent that are triggered in a particular situation and combine the knowledge and skills to adequately pursue the goals of the agent in that situation. The context that triggers the activation of a module defines the interface of the module, which can be specified declaratively, in contrast to the usual functional interpretations of such interfaces. A feature that differentiates our notion of a module from plans is that modules provide an agent with a means to focus its attention on the relevant resources it needs to handle a situation. As a result, modules can be used to control or reduce the underspecification and inherent non-determinism that is typical of agent programs. In the paper, the proposed module concept is incorporated into the agent language GOAL and illustrated by means of a simple example. [ABSTRACT FROM AUTHOR]

Several options are available to relate agent logics to computational agent systems. Among others, one can try to find useful executable fragments of an agent logic or use a model checking approach. In this paper, an alternative approach is explored based on the view that an agent logic is a program logic. Using the same starting point, one of the established agent logics, we ask instead if it is possible to construct a programming language for that agent logic. We show that the programming language and the agent logic are formally related by constructing a denotational semantics. As a result, the agent logic can be used as as a design tool to specify and verify the corresponding agent programs. In particular, we construct an agent programming language that is formally related to the KARO agent logic. The KARO logic is an agent logic that builds on top of dynamic logic. The approach is based on mapping worlds in the modal semantics for KARO onto a state-based semantics. The state-based semantics can be used to define an operational semantics for KARO programs. In this way, we obtain a computationally grounded semantics for a significant part of the KARO logic, including the operators for knowledge or beliefs, motivational attitudes and belief revision actions of a rational KARO agent. [ABSTRACT FROM AUTHOR]

We present a trust model extension that attempts to relax the assumptions that are currently taken by the majority of existing trust models: (i) proven identity of agents, (ii) repetitive interactions and (iii) similar trusting situations. The proposed approach formalizes the situation (context) and/or trusted agent identity in a multi-dimensional Identity-Context feature space, and attaches the trustworthiness evaluations to individual elements from this metric space, rather than to fixed identity tags (e.g. AIDs, addresses). Trustworthiness of the individual elements of the I-C space can be evaluated using any trust model that supports weighted aggregations and updates, allowing the integration of the mechanism with most existing work. Trust models with the proposed extension are appropriate for deployment in dynamic, ad-hoc and mobile environments, where the agent platform can't guarantee the identity of the agents and where the cryptography-based identity management techniques may be too costly due to the unreliable and costly communication. [ABSTRACT FROM AUTHOR]

Recognizing that trust states are mental states, this paper presents a formal analysis of the dynamics of trust in terms of the functional roles and representation relations for trust states. This formal analysis is done both in a logical framework and in a mathematical framework based on integral and differential equations. Furthermore, the paper presents formal specifications of a number of relevant dynamic properties of trust. The specifications provided were used to perform automated formal analysis of empirical and simulated data from two case studies, one involving two experiments with humans, and one involving simulation experiments in the context of an economic game. [ABSTRACT FROM AUTHOR]

In this paper, we propose a probabilistic framework targeting three important issues in the computation of quality and trust in decentralized systems. Specifically, our approach addresses the multi-dimensionality of quality and trust, taking into account credibility of the collected data sources for more reliable estimates, while also enabling the personalization of the computation. We use graphical models to represent peers' qualitative behaviors and exploit appropriate probabilistic learning and inference algorithms to evaluate their quality and trustworthiness based on related reports. Our implementation of the framework introduces the most typical quality models, uses the Expectation-Maximization algorithm to learn their parameters, and applies the Junction Tree algorithm to inference on them for the estimation of quality and trust. The experimental results validate the advantages of our approach: first, using an appropriate personalized quality model, our computational framework can produce good estimates, even with a sparse and incomplete recommendation data set; second, the output of our solution has well-defined semantics and useful meanings for many purposes; third, the framework is scalable in terms of performance, computation, and communication cost. Furthermore, our solution can be shown as a generalization or serve as the theoretical basis of many existing trust computational approaches. [ABSTRACT FROM AUTHOR]

Coordination systems have been used in a variety of different applications but have never performed well in large scale, faulty settings. The sheer scale and level of complexity of today's applications is enough to make the current ways of thinking about distributed systems (e.g. deterministic decisions about data organization) obsolete. All the same, computer scientists are searching for new approaches and are paying more attention to stochastic approaches that provide good solutions "most of the time". The trade-off here is that by loosening certain requirements the system ends up performing better in other fronts such as adaptiveness to failures. Adaptation is a key component to fault-tolerance and tuple distribution is the center of the fault-tolerance problem in tuple-space systems. Hence, this paper shows how the tuple distribution in Linda-like systems can be solved by using an adaptive self-organized approach à la Swarm Intelligence. The results discussed in this paper demonstrate that efficient and adaptive solutions to this problem can be achieved using simple and inexpensive approaches. [ABSTRACT FROM AUTHOR]

Coalition formation in social networks consisting of a graph of interdependent agents allows many choices of which task to select and with whom to partner in the social network. Nodes represent agents and arcs represent communication paths for requesting team formation. Teams are formed in which each agent must be connected to another agent in the team by an arc. Agents discover effective network structures by adaptation. Agents also use several strategies for selecting the task and determining when to abandon an incomplete coalition. Coalitions are finalized in one-on-one negotiation, building a working coalition incrementally. [ABSTRACT FROM AUTHOR]

Trust learning is a crucial aspect of information exchange, negotiation, and any other kind of social interaction among autonomous agents in open systems. But most current probabilistic models for computational trust learning lack the ability to take context into account when trying to predict future behavior of interacting agents. Moreover, they are not able to transfer knowledge gained in a specific context to a related context. Humans, by contrast, have proven to be especially skilled in perceiving traits like trustworthiness in such so-called initial trust situations. The same restriction applies to most multiagent learning problems. In complex scenarios most algorithms do not scale well to large state-spaces and need numerous interactions to learn. We argue that trust related scenarios are best represented in a system of relations to capture semantic knowledge. Following recent work on nonparametric Bayesian models we propose a flexible and context sensitive way to model and learn multidimensional trust values which is particularly well suited to establish trust among strangers without prior relationship. To evaluate our approach we extend a multiagent framework by allowing agents to break an agreed interaction outcome retrospectively. The results suggest that the inherent ability to discover clusters and relationships between clusters that are best supported by the data allows to make predictions about future behavior of agents especially when initial trust is involved. [ABSTRACT FROM AUTHOR]

In this paper we investigate the effect of mediated partnerships over agents' equilibrium strategies in two-sided economic search. A mediated partnership is formed when an agent acts as a mediator, establishing a partnership between a pair of agents it encountered along its search, thereby reducing the other agents' amount of search. Surprisingly, this reduction in market friction induced by mediated partnerships does not always improve market efficiency. Use of mediated partnerships changes the equilibrium strategies used by agents in two-sided search models and introduces substantial computational complexity. This computational complexity is overcome with an innovative algorithm that facilitates equilibrium calculation. [ABSTRACT FROM AUTHOR]

Past years have witnessed a growing interest in automated negotiation as a coordination mechanism for interacting agents. This paper presents a generic, problem- and domain-independent framework for argumentation-based negotiation that covers both essential agent-internal and external components relevant to automated negotiation. This framework, called Negotiation Situation Specification Scheme (N3S), is both suited as a guideline for implementing negotiation scenarios as well as integrating available approaches that address selective aspects of negotiation. In particular, N3S contributes to the state of the art in automated negotiation by identifying and relating basic argument types and negotiation stages in a structured and formal way. [ABSTRACT FROM AUTHOR]

When two or more agents interacting, their behaviors are not necessarily matching. Automated ways to overcome conflicts in the behavior of agents can make the execution of interactions more reliable. Such an alignment mechanism will reduce the necessary human intervention. This paper shows how to describe a policy for alignment, which an agent can apply when its behavior is in conflict with other agents. An extension of Petri Nets is used to capture the intended interaction of an agent in a formal way. Furthermore, a mechanism based on machine learning is implemented, to enable an agent to choose an appropriate alignment policy with collected problem information. Human intervention can reinforce certain successful policies in a given context, and can also contribute by adding completely new policies. Experiments have been conducted to test the applicability of the alignment mechanism and the main results are presented here. [ABSTRACT FROM AUTHOR]

The increasing popularity of the "business ecosystem" concept in (business) strategy reflects that it is seen as one way to cope with increasingly dynamic and complex business environments. Nevertheless, the lack of a convincing model of a business ecosystem has led to the development of software which only give organisations a partial aid whilst neglecting their need for adaptation. Research in Multi-Agent Systems has proved to be suitable for modelling interactions among disparate sort of entities such as organisations. On the other hand, natural ecosystems continue to adapt themselves to changes in their dynamic and complex environments. In this paper, we present the Dynamic Agent-based Ecosystem Model. It combines ideas from natural ecosystems and multi-agent systems for business interactions. [ABSTRACT FROM AUTHOR]

We introduce the Consensual N-Player Prisoner's Dilemma as a large-scale dilemma. We then present a framework for cooperative consensus formation in large-scale MAS under the N-Person Prisoner's Dilemma. Forming consensus is performed by demonstrating the applicability of a low-complexity physics-oriented approach to a large-scale ad hoc network problem. The framework is based on modeling cooperative MAS by a physics percolation theory. According to the model, agent-systems inherit physical properties, and therefore the evolution of the computational systems is similar to the evolution of physical systems. Specifically, we focus on the percolation theory, the emergence of self-organized criticality, and the exploitation of phase transitions. We provide a detailed low-ordered algorithm to be used by a single agent and implement this algorithm in our simulations. Via these approaches we demonstrate effective message delivery in a large-scale ad hoc network that consists of thousands of agents. [ABSTRACT FROM AUTHOR]

Coalitions are often required for multi-agent collaboration. In this research, we consider tasks that can only be completed with the combined efforts of multiple agents using approaches which are both cooperative and competitive. Often agents forming coalitions determine optimal coalitions by looking at all possibilities. This requires an exponential algorithm and is not feasible when the number of agents and tasks is large. We propose agents use a two step process of first determining the task, and secondly, the agents that will be solicited to help complete the task. We describe polynomial time heuristics for each decision. We measure four different agent types using the described heuristics. We explore diminishing choices and performance under various parameters. [ABSTRACT FROM AUTHOR]

Recent years of empirical research have collected enough evidences that for efficient markets the process of lower-wealth accumulation by capital investment is approximated by log-normal and high-wealth range by Pareto wealth distribution. This research aims to construct a simple neural network (NN) based multiagent system of heterogeneous agents' targeted to get on the efficiency frontier by combining investments to the real life index funds and nonrisky financial assets, diversifying the risk and maximizing the profits. Each agent is represented by the different stock trading strategy according to his portfolio, saving and risk aversion preferences. The goal is, following empirical evidences from the real investment markets, to find enough proofs that NN-based multiagent system, in principle, has the same fundamental properties of real investment markets described by the log-normal, Pareto wealth and Levy stock returns distributions and can be used further to simulate even more complex social phenomena. [ABSTRACT FROM AUTHOR]

This paper presents a design of high-performance agent-based intrusion detection system designed for deployment on high-speed network links. To match the speed requirements, wire-speed data acquisition layer is based on hardware-accelerated NetFlow like probe, which provides overview of current network traffic. The data is then processed by detection agents that use heterogenous anomaly detection methods. These methods are correlated by means of trust and reputation models, and the conclusions regarding the maliciousness of individual network flows is presented to the operator via one or more analysis agents, that automatically gather supplementary information about the potentially malicious traffic from remote data sources such as DNS, whois or router configurations. Presented system is designed to help the network operators efficiently identify malicious flows by automating most of the surveillance process. [ABSTRACT FROM AUTHOR]

Physically based rendering is the process of generating a 2D image from the abstract description of a 3D Scene. Despite the development of various new techniques and algorithms, the computational requirements of generating photorealistic images still do not allow to render in real time. Moreover, the configuration of good render quality parameters is very difficult and often too complex to be done by non-expert users. This paper describes a novel approach called MAgarRO (standing for "Multi-Agent AppRoach to Rendering Optimization") which utilizes principles and techniques known from the field of multi-agent systems to optimize the rendering process. Experimental results are presented which show the benefits of MAgarRO-based rendering optimization. [ABSTRACT FROM AUTHOR]

In this paper we investigate methods for analyzing the expected value of adding information in distributed task scheduling problems. As scheduling problems are NP-complete, no polynomial algorithms exist for evaluating the impact a certain constraint, or relaxing the same constraint, will have on the global problem. We present a general approach where local agents can estimate their problem tightness, or how constrained their local subproblem is. This allows these agents to immediately identify many problems which are not constrained, and will not benefit from sending or receiving further information. Next, agents use traditional machine learning methods based on their specific local problem attributes to attempt to identify which of the constrained problems will most benefit from human attention. We evaluated this approach within a distributed cTAEMS scheduling domain and found this approach was overall quite effective. [ABSTRACT FROM AUTHOR]

Traffic management nowadays is one of the key challenges for cities. One drawback of traditional approaches for traffic management is that they do not consider the different valuations of waiting-time reduction of the drivers. These valuations can differ from driver to driver, e.g., drivers who are late for their job interview have a higher valuation of reduced waiting time than individuals driving home from work routinely. This also applies to trucks with urgent load, e.g., as part of a just-in-time production chain. To overcome this problem, we propose a new mechanism for traffic control at intersections called Initial Time-Slot Auction that is valuation-aware. It relies on agent-based driver-assistance systems to allocate the right to cross an intersection. Our evaluation shows that it does yield a significantly higher overall satisfaction. [ABSTRACT FROM AUTHOR]

In this paper we study Distributed Data Mining from a Distributed Artificial Intelligence perspective. Very often, databases are very large to be mined. Then Distributed Data Mining can be used for discovering knowledge (rule sets) generated from parts of the entire training data set. This process requires cooperation and coordination between the processors because incon-sistent, incomplete and useless knowledge can be generated, since each processor uses partial data. Cooperation and coordination are important issues in Distributed Artificial Intelligence and can be accomplished with different techniques: planning (centralized, partially distributed and distributed), negotiation, reaction, etc. In this work we discuss a coordination protocol for cooperative learning agents of a MAS developed previously, comparing it conceptually with other learning systems. This cooperative process is hierarchical and works under the coordination of a manager agent. The proposed model aims to select the best rules for integration into the global model without, however, decreasing its accuracy rate. We have also done experiments comparing accuracy and complexity of the knowledge generated by the cooperative agents. [ABSTRACT FROM AUTHOR]

Recent advances in peer to peer (P2P) search algorithms have presented viable structured and unstructured approaches for full-text search. We posit that these existing approaches are each best suited for different types of queries. We present PHIRST, the first system to facilitate effective full-text search within P2P networks. PHIRST works by effectively leveraging between the relative strengths of these approaches. Similar to structured approaches, agents first publish terms within their stored documents. However, frequent terms are quickly identified and not exhaustively stored, resulting in a significantly reduction in the system's storage requirements. During query lookup, agents use unstructured searches to compensate for the lack of fully published terms. Additionally, they explicitly weigh between the costs involved with structured and unstructured approaches, allowing for a significant reduction in query costs. We evaluated the effectiveness of our approach using both real-world and artificial queries. We found that in most situations our approach yields near perfect recall. We discuss the limitations of our system, as well as possible compensatory strategies. [ABSTRACT FROM AUTHOR]

In this paper we propose a multi-agent architecture, made of co-operative information agents, where agents can share with one another their knowledge of the environment and expertise in planning for achieving goals. In particular we consider how through communication such agents can incrementally learn partial and full plans. Such information exchange is particularly useful in the case of situated agents which have diverse abilities and expertise and which have partial views of their environments. It is also useful in the case of agent systems where agents collaborate towards achieving joint or individual goals. We describe an agent model based on abductive logic programming and give detailed protocols and policies of communication. We then define formally what it means for such information exchanges to be effective, and prove results regarding termination and effectiveness of dialogues based on the formalized policies. [ABSTRACT FROM AUTHOR]

In this paper we compare state-of-the-art multi-agent reinforcement learning algorithms in a wide variety of games. We consider two types of algorithms: value iteration and policy iteration. Four characteristics are studied: initial conditions, parameter settings, convergence speed, and local versus global convergence. Global convergence is still difficult to achieve in practice, despite existing theoretical guarantees. Multiple visualizations are included to provide a comprehensive insight into the learning dynamics. [ABSTRACT FROM AUTHOR]

The concept of delegation is central to an understanding of the interactions between agents in cooperative agent problem-solving contexts. In fact, the concept of delegation offers a means for studying the formal connections between mixed-initiative problem-solving, adjustable autonomy and cooperative agent goal achievement. In this paper, we present an exploratory study of the delegation concept grounded in the context of a relatively complex multi-platform Unmanned Aerial Vehicle (UAV) catastrophe assistance scenario, where UAVs must cooperatively scan a geographic region for injured persons. We first present the scenario as a case study, showing how it is instantiated with actual UAV platforms and what a real mission implies in terms of pragmatics. We then take a step back and present a formal theory of delegation based on the use of 2APL and KARO. We then return to the scenario and use the new theory of delegation to formally specify many of the communicative interactions related to delegation used in achieving the goal of cooperative UAV scanning. The development of theory and its empirical evaluation is integrated from the start in order to ensure that the gap between this evolving theory of delegation and its actual use remains closely synchronized as the research progresses. The results presented here may be considered a first iteration of the theory and ideas. [ABSTRACT FROM AUTHOR]

Effective deployment and utilisation of limited and constrained intelligence resources — including sensors and other sources — is seen as a key issue in modern multinational coalition operations. In this talk, I will examine the application of semantic matchmaking and argumentation technologies to the management of these resources. I will show how ontologies and reasoning can be used to assign sensors and sources to meet the needs of missions, and show how argumentation can support the process of gathering and reasoning about uncertain evidence obtained from sensor probes. [ABSTRACT FROM AUTHOR]

In multi-issue negotiations, issues may be negotiated independently or not. In the latter case, the utility associated with one issue depends on the value of another. These issue dependencies give rise to more complex, non-linear utility spaces. As a consequence, the computational cost and complexity of negotiating interdependent issues is increased significantly compared to the case of independent issues. Several techniques have been proposed to deal with this increased complexity, including, for example, introducing a mediator in the negotiation setting. In this paper, we propose an alternative approach based on a weighted approximation technique to simplify the utility space. We show that given certain natural assumptions about the outcome of negotiation the application of this technique results in an outcome that closely matches with the outcome based on the original, interdependent utility structure. Moreover, using the approximated utility structure, each of the issues can be negotiated independently which ensures that the negotiation is computationally tractable. The approach is illustrated by applying and testing it in a case study. [ABSTRACT FROM AUTHOR]

Intelligent agents in the agent language 3APL are computational entities consisting of beliefs and goals which make up their mental state. In this paper, we integrate communication at the agent level into the language 3APL. Communication at the agent level consists of communication between agents of their beliefs and goals. We propose two pairs of communication primitives for agent level communication. The semantics of these primitives is based on two distinct types of reasoning: deduction and abduction. Deduction serves to derive information from a received message. Abduction serves to obtain proposals in reply to requests. We illustrate the use of these communication primitives for agents by an implementation of a multi-stage negotiation protocol. We also discuss the relation of our approach to approaches to communication based on speech act theory. [ABSTRACT FROM AUTHOR]

An important issue when defining a rule-based agent programming language is the design of interpreters for these programming languages. Since these languages are all based on some notion of rule, an interpreter must provide some means of selection from a set of such rules. We provide a concrete and intuitive ordering on rules on which this selection can be based. This ordering is inspired by the common sense metaphor of intelligent agents in Artificial Intelligence. Furthermore, we provide a language with a formal semantics for programming agent interpreters. The main idea is not to integrate this language into the agent language itself, but to provide the constructs for building interpreters at another (meta) level of specification. The operational semantics is accordingly specified at the meta level, by means of a (meta) transition system. Using this language we make a comparison between several interpreters for agent languages in the literature. [ABSTRACT FROM AUTHOR]