ACHIEVING FLEXIBLE AUTONOMY IN MULTIAGENT SYSTEMS USING CONSTRAINTS

Organizations influence many aspects of our lives. They exist for one reason: they can accomplish things that individuals cannot. While recent work in high-autonomy systems has shown that autonomy is a critical issue in artificial intelligence (AI) systems, these systems must also be able to cooperate with and rely on one another to deal with complex problems. The autonomy of such systems must be flexible, in order that agents may solve problems on their own as well as in groups. We have developed a model of distributed problem solving in which coordination of problem-solving agents is viewed as a multiagent constraint-satisfaction planning problem. This paper describes the experimental testbed that we are currently developing to facilitate the investigation of various constraint-based strategies for addressing the coordination issues inherent in cooperative distributed problem-solving domains.     

Learning Situation-Specific Coordination in Cooperative Multi-agent Systems

Achieving effective cooperation in a multi-agent system is a difficult problem for a number of reasons such as limited and possibly out-dated views of activities of other agents and uncertainty about the outcomes of interacting non-local tasks. In this paper, we present a learning system called COLLAGE, that endows the agents with the capability to learn how to choose the most appropriate coordination strategy from a set of available coordination strategies. COLLAGE relies on meta-level information about agents' problem solving situations to guide them towards a suitable choice for a coordination strategy. We present empirical results that strongly indicate the effectiveness of the learning algorithm.     

Guest Editors' Introduction: Planning with Templates

Templates are standard operating procedures that can be used for solving typical problems and as a starting point for solving novel problems. These structures contain relevant variables (and required activities) and current variable values (or specific activities) that affect a problem or have been chosen in a problem-solving instance. With templates you can configure domain-independent planning algorithms, making them applicable to many different problem domains. By making the template explicit for the user in the form of a GUI, you can facilitate mixed-initiative, user-centric systems that help maintain awareness of complex and dynamic situations, share information across the network, and solve problems incrementally and iteratively.     

Negotiation as a metaphor for distributed problem solving

We describe the concept of distributed problem solving and define it as the cooperative solution of problems by a decentralized and loosely coupled collection of problem solvers. This approach to problem solving offers the promise of increased performance and provides a useful medium for exploring and developing new problem-solving techniques.We present a framework called the contract net that specifies communication and control in a distributed problem solver. Task distribution is viewed as an interactive process, a discussion carried on between a node with a task to be executed and a group of nodes that may be able to execute the task. We describe the kinds of information that must be passed between nodes during the discussion in order to obtain effective problem-solving behavior. This discussion is the origin of the negotiation metaphor: Task distribution is viewed as a form of contract negotiation.We emphasize that protocols for distributed problem solving should help determine the content of the information transmitted, rather than simply provide a means of sending bits from one node to another.The use of the contract net framework is demonstrated in the solution of a simulated problem in area surveillance, of the sort encountered in ship or air traffic control. We discuss the mode of operation of a distributed sensing system, a network of nodes extending throughout a relatively large geographic area, whose primary aim is the formation of a dynamic map of traffic in the area.From the results of this preliminary study we abstract features of the framework applicable to problem solving in general, examining in particular transfer of control. Comparisons with planner, conniver, hearsay-ii, and pup6 are used to demonstrate that negotiation—the two-way transfer of information—is a natural extension to the transfer of control mechanisms used in earlier problem-solving systems.     

Reflections on the Nature of Multi-Agent Coordination and Its Implications for an Agent Architecture

The development of enabling infrastructure for the next generation of multi-agent systems consisting of large numbers of agents and operating in open environments is one of the key challenges for the multi-agent community.Current infrastructure support does not materially assist in the development of sophisticated agent coordination strategies. It is the need for and the development of such a high-level support structure that will be the focus of this paper. A domain-independent (generic) agent architecture is proposed that wraps around an agent's problem-solving component in order to make problem solving responsive to real-time constraints, available network resources, and the need to coordinate—both in the large and small—with problem-solving activities of other agents. This architecture contains five components, local agent scheduling, multi-agent coordination, organizational design, detection and diagnosis, and on-line learning, that are designed to interact so that a range of different situation-specific coordination strategies can be implemented and adapted as the situation evolves. The presentation of this architecture is followed by a more detailed discussion on the interaction among these components and the research questions that need to be answered to understand the appropriateness of this architecture for the next generation of multi-agent systems.     

Gaze-augmented interaction improves problem-solving: new evidence from verbal protocols

When interacting with a problem on a computer screen, users need to select from a range of strategies that the user interface supports. Previous research shows that the problem-solving user interface plays a significant role in the strategy selection, such as by making recall unnecessary or by facilitating planning. What is not very well understood is how the interaction devices, themselves, affect problem-solving strategies. We conducted an experiment in which users interacted with a problem using either a computer mouse or gaze-augmented input. We analysed the qualitative differences in the problem-solving process by investigating the content of verbal protocols. We present a new evidence that shows that the gaze-based interaction is characterised by an increased cognitive processing of thecurrently attended information, leading to better plan development and improved problem-solving strategies. The findings have implications on the design of future gaze-aware problem-solving user interfaces.     

Individual learning of coordination knowledge

Abstract. Social agents, both human and computational, inhabiting a world containing multiple active agents, need to coordinate their activities. This is because agents share resources, and without proper coordination or ‘rules of the road’, everybody will be interfering with the plans of others. As such, we need coordination schemes that allow agents to effectively achieve local goals without adversely affecting the problem-solving capabilities of other agents. Researchers in the field of Distributed Artificial Intelligence (DAI) have developed a variety of coordination schemes under different assumptions about agent capabilities and relationships. Whereas some of these researchers have been motivated by human cognitive biases, others have approached it as an engineering problem of designing the most effective coordination architecture or protocol. We evaluate individual and concurrent learning by multiple, autonomous agents as a means for acquiring coordination knowledge. We show that a uniform reinforcement learning algorithm suffices as a coordination mechanism in both cooperative and adversarial situations. Using a number of multi-agent learning scenarios with both tight and loose coupling between agents and with immediate as well as delayed feedback, we demonstrate that agents can consistently develop effective policies to coordinate their actions without explicit information sharing. We demonstrate the viabilityof using both the Q-learning algorithm and genetic algorithm based classifier systems with different pay-off schemes, namely the bucket brigade algorithm (BBA) and the profit sharing plan (PSP), for developing agent coordination on two different multi-agent domains. In addition, we show that a semi-random scheme for action selection is preferable to the more traditional fitness proportionate selection scheme used in classifier systems.     

An auto-scoring mechanism for evaluating problem-solving ability in a web-based learning environment

The rapid development of computer and network technologies has attracted researchers to investigate strategies for and the effects of applying information technologies in learning activities; simultaneously, learning environments have been developed to record the learning portfolios of students seeking web information for problem-solving. Although previous research has demonstrated the benefits of applying information technologies to learning activities, the difficulties in doing so have also been revealed. One of the major difficulties is the lack of a mechanism to assist teachers in evaluating the problem-solving ability of the students, such that constructive suggestions can be given to the students, and tutoring strategies can be improved accordingly.     

A distributed control architecture for an autonomous mobile robot

This paper describes a Distributed Control Architecture for an autonomous mobile robot. We start by characterizing the Conceptual Levels into which the various problem-solving activities of a mobile robot can be classified. In sequence, we discuss a Distributed Control System that provides scheduling and coordination of multiple concurrent activities on a mobile robot. Multiple Expert Modules are responsible for the various tasks and communicate through messages and over a Blackboard. As a testbed, the architecture of a specific system for Sonar-Based Mapping and Navigation is presented, and a distributed implementation is described.     

视觉化协同超媒体及其在合作学习中的应用

在分布式合作学习环境中,学生和老师需要一个共享的信息空间去与学习材料交互作用,并从中知晓合作者的存在及其活动.学生和老师还需要一些协同和通信手段去协调他们共同的活动.为此,提出了一种视觉化协同超媒体技术.它可以提供所需的共享信息空间.在此空间中,形象的超媒体对象既可以表示学习材料(学习内容),又可以表示学习过程(学习方法).这种协同超媒体集成了灵活的过程支持、通信渠道以及文件处理功能.因而,它可以支持学生(和老师)按一定的学习方法并通过文字或声音渠道去协调他们共同的学习活动.这种协同超媒体可以通过万维网(浏览器)访问.为了揭示这种技术的实用性,还给出了一个案例和一些用户体验.     

Documentarisation Processes in Documents for Action (DofA): The Status of Annotations and Associated Cooperation Technologies

In this paper, we focus on situations where documents serve to coordinate the work of a distributed collective engaged in common goal-directed activities. After defining the concept of semiotic products as resulting from symbolic communicational transactions, we present some coordination strategies which can be used to compensate for the spatio-socio-temporal distribution typical of these transactions. Among these strategies, it is proposed to study in detail the documentarisation strategy, which makes the material substrate mediating the transactions relatively durable and endows it with attributes making its further use possible. In our study of documentarisation processes, several novel concepts are introduced and used to describe Documents for Action (DofA), their characteristics and the conditions that should be respected for correctly annotating them.     

Robust finite-time containment control for high-order multi-agent systems with matched uncertainties under directed communication graphs

In this paper, we study the robust finite-time containment control problem for a class of high-order uncertain nonlinear multi-agent systems modelled as high-order integrator systems with bounded matched uncertainties. When relative state information between neighbouring agents is available, an observer-based distributed controller is proposed for each follower using the sliding mode control technique which solves the finite-time containment control problem under general directed communication graphs. When only relative output information is available, robust exact differentiators and high-order sliding-mode controllers are employed together with the distributed finite-time observers. It is shown that robust finite-time containment control can still be achieved in this situation. An application in the coordination of multiple non-holonomic mobile robots is used as an example to illustrate the effectiveness of the proposed control strategies.     

Learning task models in ill-defined domain using an hybrid knowledge discovery framework

Domain experts should provide Intelligent Tutoring Systems (ITS) with relevant domain knowledge that enable it to guide the learner during problem-solving learning activities. However, for ill-defined domains this knowledge is hard to define explicitly. Our hypothesis is that knowledge discovery (KD) techniques can be used to extract problem-solving task models from the recorded usage of expert, intermediate and novice learners. This paper proposes a procedural-knowledge acquisition framework based on a combination of sequential pattern mining and association rules discovery techniques. The framework has been implemented and is used to discover new meta-knowledge and rules in a given domain which then extend domain knowledge and serve as problem space, allowing the Intelligent Tutoring System to guide learners in problem-solving situations. Preliminary experiments have been conducted using the framework as an alternative to a path-planning problem solver in CanadarmTutor.     

Team Coordination in Escalating Situations: An Empirical Study Using Mid‐Fidelity Simulation

The performance of teams, with different levels of domain and crisis management experience, managing unexpected and escalating situations was observed using a mid‐fidelity ship‐bridge simulation and analysed by applying the central concepts of joint activity coordination as well as Woods's theory building on data overload. The coordination strategies used by the teams were evaluated by applying coordination process indicators and the concept of control. The paper discusses how different aspects of team coordination in unexpected and escalating situations, e.g. that teams that maintain a high level of control in escalating situations, avoid or minimize the effects of data overload by using explicit and agreed‐upon goals rather than sharing as much incoming information as possible. The results presented in this paper show the benefits of applying a broad set of theoretical concepts to shed light on the actual demands that escalating situations pose on people's data processing capacities and processes. It also provides guidance on the successful performance of teams in such situations and thus support for the development of successful strategies for their management.     

企业营销信息系统的研究

营销信息系统将企业与市场环境紧密联系起来，是成功营销实践的基础。本文分析了营销信息系统的功能需求和信息需求，认为营销信息系统需要满足协作，智能和分布式处理，在此基础上，提出了基于多智能体的求解模型，由交互层、问题分解层和问题求解层构成的，以任务智能体为主构成系统的求解能力。     

A hybrid approach to promoting students’ web-based problem-solving competence and learning attitude

Fostering problem-solving abilities has long been recognized as an important issue in education; however, past studies have shown that it is difficult and challenging to find effective learning strategies or tools for improving students’ problem-solving abilities. To cope with this problem, in this study, a hybrid approach that integrates the cognitive apprenticeship model with the collaborative learning strategy is proposed for conducting web-based problem-solving activities. Students’ problem-solving performance is examined in such a hybrid learning context. Furthermore, past studies indicate that cognitive load could affect learners’ performance; thus, the influence of cognitive load on students’ problem-solving effectiveness with this new approach is investigated in depth. The experimental results show that middle- and low-achievement students in the experimental group gained significant benefits from the hybrid approach in comparison with those who learned with the traditional approach. Accordingly, a discussion of how to accommodate the needs of different learning ability groups is provided.     

Supporting cognitive processing with spatial information presentations in virtual environments

While it has been suggested that immersive virtual environments could provide benefits for educational applications, few studies have formally evaluated how the enhanced perceptual displays of such systems might improve learning. Using simplified memorization and problem-solving tasks as representative approximations of more advanced types of learning, we are investigating the effects of providing supplemental spatial information on the performance of learning-based activities within virtual environments. We performed two experiments to investigate whether users can take advantage of a spatial information presentation to improve performance on cognitive processing activities. In both experiments, information was presented either directly in front of the participant, at a single location, or wrapped around the participant along the walls of a surround display. In our first experiment, we measured memory scores and analyzed participant strategies for a memorization and recall task. In addition to comparing spatial and non-spatial presentations, we also varied field of view and background imagery. The results showed that the spatial presentation caused significantly better memory scores. Additionally, a significant interaction between background landmarks and presentation style showed that participants used more visualization strategies during the memorization task when background landmarks were shown with spatial presentations. To investigate whether the advantages of spatial information presentation extend beyond memorization to higher level cognitive activities, our second experiment employed a puzzle-like task that required critical thinking using the presented information. Focusing only on the effects of spatial presentations, this experiment measured task performance and mental workload. The results indicate that no performance improvements or mental workload reductions were gained from the spatial presentation method compared with a non-spatial layout for our problem-solving task. The results of these two experiments suggest that supplemental spatial information can affect mental strategies and support performance improvements for cognitive processing and learning-based activities. However, the effectiveness of spatial presentations is dependent on the nature of the task and a meaningful use of space and may require practice with spatial strategies.     

利用问题求解理论来研究交互式复杂信息的可视分析行为

面对大数据的挑战,力图将人的推理能力和计算系统的数据处理能力相结合的交 互式可视分析研究变得愈发重要。然而目前仍缺乏有效的认知理论来指导面向复杂信息的可视 分析系统的设计,诸如意义构建等现有的理论框架通常着眼于分析行为的外在特征,未能对此 类行为的内在认知机理进行深入研究。因此提出将问题求解作为一种理论框架来解释交互可视 分析行为的基本认知活动,并建议从非良构问题的角度来描述可视分析过程中用户所面临的主 要挑战,还从问题表征及问题求解策略等角度分析了可视分析系统对分析行为的影响。本研究 在理论上,将认知心理学领域的问题求解理论引入到交互可视分析行为的研究中,该方法对设 计面向复杂信息分析的其他类型交互系统也有启示作用;在实践层面上,从问题求解的支持角 度探索了可视分析系统的设计和评估问题。     

基于多Agent的组合模型链分布式求解算法及实现

在决策支持系统中,组合模型往往是解决复杂决策问题的有效途径,但无论采用何种模型组合技术,随着模型库模型的增多,模型组合过程中生成的模型链数量都会快速增加,这样势必造成求解速度上的瓶颈。针对这种情况,文章提出了一种基于多Agent的组合模型分布式求解模型,充分利用分布式技术快速对组合模型实现并行求解,从而提高模型组合求解的效率,保证决策的实时性。