Using temporal-difference learning for multi-agent bargaining

This research treats a bargaining process as a Markov decision process, in which a bargaining agent’s goal is to learn the optimal policy that maximizes the total rewards it receives over the process. Reinforcement learning is an effective method for agents to learn how to determine actions for any time steps in a Markov decision process. Temporal-difference (TD) learning is a fundamental method for solving the reinforcement learning problem, and it can tackle the temporal credit assignment problem. This research designs agents that apply TD-based reinforcement learning to deal with online bilateral bargaining with incomplete information. This research further evaluates the agents’ bargaining performance in terms of the average payoff and settlement rate. The results show that agents using TD-based reinforcement learning are able to achieve good bargaining performance. This learning approach is sufficiently robust and convenient, hence it is suitable for online automated bargaining in electronic commerce.     

RMI在海上搜救智能决策支持系统中的应用

对分布式人工智能领域中的重点问题——多Agent系统（MAS）的通信方式进行分析,论述基于MAS的海上搜救智能决策支持系统的通信机制以及系统的底层实现方法,介绍远程方法调用（RMI）在基于MAS的海上搜救智能决策支持系统中的具体应用,对RMI通信的主要优势进行总结。     

Model-driven engineering techniques for the development of multi-agent systems

Model-driven engineering (MDE), implicitly based upon meta-model principles, is gaining more and more attention in software systems due to its inherent benefits. Its use normally improves the quality of the developed systems in terms of productivity, portability, inter-operability and maintenance. Therefore, its exploitation for the development of multi-agent systems (MAS) emerges in a natural way. In this paper, agent-oriented software development (AOSD) and MDE paradigms are fully integrated for the development of MAS. Meta-modeling techniques are explicitly used to speed up several phases of the process. The Prometheus methodology is used for the purpose of validating the proposal. The meta-object facility (MOF) architecture is used as a guideline for developing a MAS editor according to the language provided by Prometheus methodology. Firstly, an Ecore meta-model for Prometheus language is developed. Ecore is a powerful tool for designing model-driven architectures (MDA). Next, facilities provided by the Graphical Modeling Framework (GMF) are used to generate the graphical editor. It offers support to develop agent models conform to the meta-model specified. Afterwards, it is also described how an agent code generator can be developed. In this way, code is automatically generated using as input the model specified with the graphical editor. A case of study validates the method put in practice for the development of a multi-agent surveillance system.     

基于CPN与Aalaadin元模型的MAS建模

对于MAS而言,传统的UML已经无法满足建模的需求,必须找到一种更好的方式来对MAS建模,这里对UML进行了研究和扩展至AUML,以达到对建模的需求,还使用了CPN与Aalaadin元模型来弥补AUML的不足,CPN与Aalaadin元模型也能很好的互为补充,极大提高了MAS模型的可行性.     

Artifacts in the A&;A meta-model for multi-agent systems

In this article we focus on the notion of artifact for agents in multi-agent systems (MAS) as a basis for a new meta-model promoting the modelling and engineering of agent societies and MAS environment as first-class entities. Its conceptual foundations lay upon theories and results coming from computational sciences as well as from organisational and cognitive sciences, psychology, computer supported cooperative work (CSCW), anthropology and ethology. In the resulting agents & artifacts (A&A) meta-model, agents are the (pro-)active entities in charge of the goals/tasks that altogether build up the whole MAS behaviour, whereas artifacts are the reactive entities providing the services and functions that make individual agents work together in a MAS, and that shape agent environment according to the MAS needs. After presenting the scientific background, we define the notions of artifact in the A&A meta-model, discuss how it affects the notion of intelligence in MAS, and show its application to a number of agent-related research fields.     

Predicting the Expected Behavior of Agents that Learn About Agents: The CLRI Framework

We describe a framework and equations used to model and predict the behavior of multi-agent systems (MASs) with learning agents. A difference equation is used for calculating the progression of an agent's error in its decision function, thereby telling us how the agent is expected to fare in the MAS. The equation relies on parameters which capture the agent's learning abilities, such as its change rate, learning rate and retention rate, as well as relevant aspects of the MAS such as the impact that agents have on each other. We validate the framework with experimental results using reinforcement learning agents in a market system, as well as with other experimental results gathered from the AI literature. Finally, we use PAC-theory to show how to calculate bounds on the values of the learning parameters.     

A general knowledge mediation infrastructure for multi-agent systems

Intelligent decision making needs to be equipped with broader knowledge in order to enhance the decision quality. Knowledge for decision making can be categorized as domain specific and general. Applying domain knowledge in intelligent systems is not new, but applying general knowledge to support business decision making is a possible way to obtain an edge over competitors. For this reason, the paper focuses primarily on designing a general knowledge mediation infrastructure (GKMI) which supports the use of general knowledge from multiple heterogeneous sources, and provides an unified access point for typical multi-agent systems (MAS) to access that knowledge. The finite state automaton (FSA) is used to model and analyze the commonsense inference ability of GKMI. By carrying out two use cases of GKMI for MAS development and operation the effectiveness of this infrastructure is examined.     

A decision support system for cooperative transportation planning: Design,implementation, and performance assessment

In this paper, we describe a decision support system for cooperative transportation planning in the German food industry where several manufacturing companies share their fleets to reduce transportation costs. Besides using vehicles of their fleets, there are different outsourcing options offered by logistics service providers, but these are much more expensive. The decision-making kernel of the decision support system is implemented as a multi-agent-system (MAS). The kernel provides a distributed hierarchical algorithm for cooperative transportation planning and an on-line data layer that contains all the information for decision making. We sketch the distributed hierarchical transportation planning algorithm and identity the required software agents. The MAS interacts via web services with a commercial tour planning system that persistently stores the resulting tour plans, orders, and customer data. Moreover, the tour planning system is used to offer graphical user interfaces to interact with the users. The data layer is updated by order and customer data from the ERP systems of the different manufacturing companies. We describe the architecture and the implementation of the MAS and the overall coupling framework. Furthermore, we discuss the simulation-based performance assessment of the resulting decision support system when the system is applied in a rolling horizon setting and present some computational results. The results demonstrate that the MAS approach is appropriate for the cooperative transportation planning domain.     

Formal ReSpecT in the A&A Perspective

Coordination models and languages have found a new course in the context of MAS (multiagent systems). By re-interpreting results in terms of agent-oriented abstractions, new conceptual spaces are found, which extend the reach of coordination techniques far beyond their original scope. This is for instance the case of coordination media, when recasted in terms of coordination artifacts in the MAS context.In this paper, we take the well-established ReSpecT language for programming tuple centre behaviour, and adopt the A&A (agents and artifacts) meta-model as a perspective to reinterpret, revise, extend and complete it. A formal model of the so-called A&A ReSpecT language is presented, along with an example illustrating its use for MAS coordination.     

AGR元模型到组织抽象模型的建模方法

将目标驱动思想引入AGR中,提出一种从AGR元模型到多Agent系统（MAs）组织抽象模型的建模方法。对AGR进行综述,给出该方法的具体步骤和形式化过程。应用于某智能故障诊断系统的结果表明,该建模方法有利于设计人员直接利用AGR元模型构建MAS组织抽象模型,可适用于各种AGR扩展。     

Dynamic decision support framework for production scheduling using a combined genetic algorithm and multiagent model

Due to the dynamic nature, complexity, and interactivity of production scheduling in an actual business environment, suitable combined and hybrid methods are necessary. This paper takes prefabricated concrete components as an example and develops the dynamic decision support framework based on a genetic algorithm and multiagent system (MAS) to optimize and simulate the production scheduling. First, a multiobjective genetic algorithm is integrated into the MAS for preliminary optimization and a series of near‐optimal solutions are obtained. Subsequently, considering the resource constraints and uncertainties, the MAS is used to simulate complex real‐world production environments. Considering the different types of uncertainty factors, the paper proposes the corresponding dynamic scheduling method and uses MAS to generate the optimal production schedule. Finally, a practical prefabricated construction case is used to validate the proposed model. The results show that the model can effectively address the occurrence of uncertain events and can provide dynamic decision support for production scheduling.     

Strategic induction of decision trees

An algorithm for decision-tree induction is presented in which attribute selection is based on the evidence-gathering strategies used by doctors in sequential diagnosis. Since the attribute selected by the algorithm at a given node is often the best attribute according to the Quinlan's information gain criterion, the decision tree it induces is often identical to the ID3 tree when the number of attributes is small. In problem-solving applications of the induced decision tree, an advantage of the approach is that the relevance of a selected attribute or test can be explained in strategic terms. An implementation of the algorithm in an environment providing integrated support for incremental learning, problem solving and explanation is presented.     

Negotiation in a multi-agent system for construction claims negotiation

Negotiation is an important approach for agents to co-operate and reach agreement in multiagent systems (MAS). Different negotiation theories and models have been deployed in a variety of applications. This paper is concerned with the applicability of these theories to the domain of agent-based construction claims negotiation. The peculiarities of this domain are highlighted and the approach adopted in the development of a multi-agent system for construction claims negotiation (MASCOT) described. Of particular interest is the integration of Zeuthen's bargaining model with a Bayesian learning mechanism, which addresses the characeristics of the construction claims negotiation. Examples are presented to demonstrate the impact of various negotiation approaches on the conduct and outcome of construction claims negotiations.     

Exploring the Role of Management Accounting Systems in Strategic Sensemaking

Abstract

Management accounting systems (MAS) provide an information context that is required for strategic sensemaking. Research from the interpretive perspective of information processing suggests some mechanisms how MAS can contribute to strategic sensemaking, although it neglects the relationship between MAS use and MAS dimensions. The systems-structural perspective identified some important MAS dimensions, but it does not explain how these dimensions contribute to strategic sensemaking. The objective of this study is to explore the role of MAS in sensemaking and how MAS dimensions contribute to this role. Based on 30 interviews with top and middle managers from 7 large companies we suggest a set of MAS dimensions that relates to sensemaking.     

一种新的多智能体Q学习算法

针对非确定马尔可夫环境下的多智能体系统,提出了一种新的多智能体Q学习算法.算法中通过对联合动作的统计来学习其它智能体的行为策略,并利用智能体策略向量的全概率分布保证了对联合最优动作的选择. 同时对算法的收敛性和学习性能进行了分析.该算法在多智能体系统RoboCup中的应用进一步表明了算法的有效性与泛化能力.     

基于Agent 方法体系的ODSS 研究

通过组织与多Ａｇｅｎｔ系统的比较,指出组织问题的研究与多Ａｇｅｎｔ系统可以相互借鉴;对组织决策进行分析,指出组织知识、组织智能对组织决策的重要性;在此基础上,提出在组织环境下基于多Ａｇｅｎｔ系统的ＯＤＳＳ层次模型。     

Integrating and navigating engineering design decision-related knowledge using decision knowledge graph

Designers are usually facing a problem of finding information from a huge amount of unstructured textual documents in order to prepare for a decision to be made. The major challenge is that knowledge embedded in the textual documents are difficult to search at a semantic level and therefore not ready to support decisions in a timely manner. To address this challenge, in this paper we propose a knowledge-graph-based method for integrating and navigating decision-related knowledge in engineering design. The presented method is based on a meta-model of decision knowledge graph (mDKG) that is grounded in the compromise Decision Support Problem (cDSP) construct which is used by designers as a means to formulate design decisions linguistically and mathematically. Based on the mDKG, we propose a procedure for automatically converting word-based cDSPs to knowledge graph through natural language processing, and a procedure for rapidly and accurately navigating decision-related knowledge through divergence and convergence processes. The knowledge-graph-based method is verified using the textual data from the supply chain design domain. Results show that our method has better performance than the conventional keyword-based searching method in terms of both effectiveness and efficiency in finding the target knowledge.     

Information and modeling resources for decision support in global environments

Decision support through information and modeling resources is crucial for strategic decision making in a global environment. This paper describes how decision support systems (DSS) were developed, implemented, and utilized in an organization to evaluate strategic options for foreign direct investments in manufacturing facilities. The paper shows how the process of decision support was implemented and integrated into the company's long range strategic and annual profit planning processes. The approach to decision support can be adopted by other firms to aid international investment planning and operations.     

Knowledge Exchange in Strategic Alliances: Learning in Tension

The proliferation of corporate strategic alliances is explained by the opportunities this provides for the exchange of knowledge and more rapid learning than any other factor. Exploiting complementarities among products and services, strategic alliances enable value creation by capturing the benefits from leveraging knowledge, and discovering complementarities among technologies, and among the activities of the participants. This paper explores the relevance of factors, which may influence the relationship involving the imbalances of internal tensions and alliance instabilities. It is assumed that the objective in a strategic alliance partnership is to maintain the collaborative relationship and to prevent unplanned alliance dissolution. Factors such as availability of resources, bargaining power, alliance type, alliances with specific goals and stages of industry life cycles, and changing market conditions can influence internal tensions and therefore alliance stability. This article argues that alliance partners should balance the conflicting forces to maintain the collaborative knowledge creating and learning relationship.