A Two-Stage Online Approach for Collaborative Multi-Agent Planning under Uncertainty

In a team of multiple agents, the pursuance of a common goal is a defining characteristic. Since agents may have different capabilities, and effects of actions may be uncertain, a common goal can generally only be achieved through a careful cooperation between the different agents. In this work, we propose a novel two-stage planner that combines online planning at both team level and individual level through a subgoal delegation scheme. The proposal brings the advantages of online planning approaches to the multi-agent setting. A number of modifications are made to a classical UCT approximate algorithm to (i) adapt it to the application domains considered, (ii) reduce the branching factor in the underlying search process, and (iii) effectively manage uncertain information of action effects by using information fusion mechanisms. The proposed online multi-agent planner reduces the cost of planning and decreases the temporal cost of reaching a goal, while significantly increasing the chance of success of achieving the common goal. 

A New Technique for Studying Vapour–liquid Equilibria of Multi-Component Systems

A Fourier transform infrared gas-phase method is described herein and capable of deriving the vapour pressure of each pure component of a poorly volatile mixture and determining the relative vapour phase composition for each system. The performance of the present method has been validated using two standards (naphthalene and ferrocene), and a Raoult’s plot surface of a ternary system is reported as proof-of-principle. This technique is ideal for studying solutions comprising two, three, or more organic compounds dissolved in ionic liquids as they have no measurable vapour pressures.

A Multi-Agent Simulation of Retail Management Practices

We apply Agent-Based Modeling and Simulation (ABMS) to investigate a set of problems in a retail context. Specifically, we are working to understand the relationship between human resource management practices and retail productivity. Despite the fact we are working within a relatively novel and complex domain, it is clear that intelligent agents do offer potential for developing organizational capabilities in the future. Our multi-disciplinary research team has worked with a UK department store to collect data and capture perceptions about operations from actors within departments. Based on this case study work, we have built a simulator that we present in this paper. We then use the simulator to gather empirical evidence regarding two specific management practices: empowerment and employee development.

A Hardware Task-Graph Scheduler for Reconfigurable Multi-tasking Systems

Reconfigurable hardware can be used to build a multitasking system where tasks are assigned to HW resources at run-time according to the requirements of the running applications. These tasks are frequently represented as direct acyclic graphs and their execution is typically controlled by an embedded processor that schedules the graph execution. In order to improve the efficiency of the system, the scheduler can apply prefetch and reuse techniques that can greatly reduce the reconfiguration latencies. For an embedded processor all these computations represent a heavy computational load that can significantly reduce the system performance. To overcome this problem we have implemented a HW scheduler using reconfigurable resources. In addition we have implemented both prefetch and replacement techniques that obtain as good results as previous complex SW approaches, while demanding just a few clock cycles to carry out the computations. We consider that the HW cost of the system (in our experiments 3% of a Virtex-II PRO xc2vp30 FPGA) is affordable taking into account the great efficiency of the techniques applied to hide the reconfiguration latency and the negligible run-time penalty introduced by the scheduler computations.

A Task-Graph Execution Manager for Reconfigurable Multi-tasking Systems

Reconfigurable hardware can be used to build multi tasking systems that dynamically adapt themselves to the requirements of the running applications. This is especially useful in embedded systems, since the available resources are very limited and the reconfigurable hardware can be reused for different applications. In these systems computations are frequently represented as task graphs that are executed taking into account their internal dependencies and the task schedule. The management of the task graph execution is critical for the system performance. In this regard, we have developed two dif erent versions, a software module and a hardware architecture, of a generic task-graph execution manager for reconfigurable multi-tasking systems. The second version reduces the run-time management overheads by almost two orders of magnitude. Hence it is especially suitable for systems with exigent timing constraints. Both versions include specific support to optimize the reconfiguration process.

Social Awareness Concepts to Support Social Computing

Our social life is characterised by norms that manifest as attitudinal and behavioural uniformities among people. With greater awareness about our social context, we can interact more efficiently. Any theory or account of human interaction that fails to include social concepts could be suggested to lack a critical element. This paper identifies social concepts that need to be supported by future context-aware systems. It discusses the limitations of existing context-aware and Multi-Agent Systems (MAS) to support social psychology theories related to the identification and membership of social groups. We argue thatsocial norms are among the core modeling concepts that future context-aware systems need to capture with the view to support and enhance social interactions. The social concepts identified in this paper could be used to simulate agent interactions imbued with social norms or use ICT to facilitate, assist or enhance social interactions. They also could be used in virtual communities modeling where the awareness of a community as well as the process of joining and exiting a community are important.

Towards executable specification : combining i* and AgentSpeak(L).

Agent-oriented conceptual modelling (AoCM) approaches in Requirements Engineering (RE) have received considerable attention recently. Semi-formal modeling frameworks such as i* assist analysts in requirements elicitation and reasoning of early-phase RE. AgentSpeak(L) is a widely accepted agent programming language. The Strategic Rationale (SR) model of the i* framework naturally lends itself to AgentSpeak(L) programs. Furthermore, the Strategic Dependency (SD) component of the i* framework prescribes the interaction between the agents in a multi-agent environment. This paper proposes a formal methodology for transforming a SR model to an AgentS- peak(L) agent. The constructed AgentSpeak(L) agents will then form the essential components of a multi-agent system, MAS.

Infinite-horizon policy-gradient estimation

Gradient-based approaches to direct policy search in reinforcement learning have received much recent attention as a means to solve problems of partial observability and to avoid some of the problems associated with policy degradation in value-function methods. In this paper we introduce GPOMDP, a simulation-based algorithm for generating a biased estimate of the gradient of the average reward in Partially Observable Markov Decision Processes (POMDPs) controlled by parameterized stochastic policies. A similar algorithm was proposed by Kimura, Yamamura, and Kobayashi (1995). The algorithm's chief advantages are that it requires storage of only twice the number of policy parameters, uses one free parameter β ∈ [0,1) (which has a natural interpretation in terms of bias-variance trade-off), and requires no knowledge of the underlying state. We prove convergence of GPOMDP, and show how the correct choice of the parameter β is related to the mixing time of the controlled POMDP. We briefly describe extensions of GPOMDP to controlled Markov chains, continuous state, observation and control spaces, multiple-agents, higher-order derivatives, and a version for training stochastic policies with internal states. In a companion paper (Baxter, Bartlett, & Weaver, 2001) we show how the gradient estimates generated by GPOMDP can be used in both a traditional stochastic gradient algorithm and a conjugate-gradient procedure to find local optima of the average reward. ©2001 AI Access Foundation and Morgan Kaufmann Publishers. All rights reserved.

General theory of stigmergy : modelling stigma semantics

Stigmergy is a biological term used when discussing a sub-set of insect swarm-behaviour describing the apparent organisation seen during their activities. Stigmergy describes a communication mechanism based on environment-mediated signals which trigger responses among the insects. This phenomenon is demonstrated in the behavior of ants and their food gathering process when following pheromone trails, where the pheromones are a form of environment-mediated communication. What is interesting with this phenomenon is that highly organized societies are achieved without an apparent management structure. Stigmergy is also observed in human environments, both natural and engineered. It is implicit in the Web where sites provide a virtual environment supporting coordinative contributions. Researchers in varying disciplines appreciate the power of this phenomenon and have studied how to exploit it. As stigmergy becomes more widely researched we see its definition mutate as papers citing original work become referenced themselves. Each paper interprets these works in ways very specific to the research being conducted. Our own research aims to better understand what improves the collaborative function of a Web site when exploiting the phenomenon. However when researching stigmergy to develop our understanding we discover a lack of a standardized and abstract model for the phenomenon. Papers frequently cited the same generic descriptions before becoming intimately focused on formal specifications of an algorithm, or esoteric discussions regarding sub-facets of the topic. None provide a holistic and macro-level view to model and standardize the nomenclature. This paper provides a content analysis of influential literature documenting the numerous theoretical and experimental papers that have focused on stigmergy. We establish that stigmergy is a phenomenon that transcends the insect world and is more than just a metaphor when applied to the human world. We present from our own research our general theory and abstract model of semantics of stigma in stigmergy. We hope our model will clarify the nuances of the phenomenon into a useful road-map, and standardise vocabulary that we witness becoming confused and divergent. Furthermore, this paper documents the analysis on which we base our next paper: Special Theory of Stigmergy: A Design Pattern for Web 2.0 Collaboration.

A Monte-Carlo tree search in argumentation

Monte-Carlo Tree Search (MCTS) is a heuristic to search in large trees. We apply it to argumentative puzzles where MCTS pursues the best argumentation with respect to a set of arguments to be argued. To make our ideas as widely applicable as possible, we integrate MCTS to an abstract setting for argumentation where the content of arguments is left unspecified. Experimental results show the pertinence of this integration for learning argumentations by comparing it with a basic reinforcement learning.

Determining Top K Nodes in Social Networks using the Shapley Value

In this paper, we consider the problem of selecting, for any given positive integer k, the top-k nodes in a social network, based on a certain measure appropriate for the social network. This problem is relevant in many settings such as analysis of co-authorship networks, diffusion of information, viral marketing, etc. However, in most situations, this problem turns out to be NP-hard. The existing approaches for solving this problem are based on approximation algorithms and assume that the objective function is sub-modular. In this paper, we propose a novel and intuitive algorithm based on the Shapley value, for efficiently computing an approximate solution to this problem. Our proposed algorithm does not use the sub-modularity of the underlying objective function and hence it is a general approach. We demonstrate the efficacy of the algorithm using a co-authorship data set from e-print arXiv (www.arxiv.org), having 8361 authors.

Integrando multiagentes em ambientes 3D: um serious game para estimulação cognitiva

Sistemas Multiagentes estão recebendo cada vez mais a atenção de pesquisadores e desenvolvedores de jogos virtuais. O uso de agentes permite controlar o desempenho do usuário, adaptando a interface e alterando automaticamente o nível de dificuldade das tarefas. Este trabalho descreve uma estratégia de integração de sistemas multiagentes e ambientes virtuais tridimensionais e exemplifica a viabilidade dessa integração através do desenvolvimento de um jogo com características de Serious game. Este jogo visa estimular as funções cognitivas, tais como atenção e memória e é voltado para pessoas portadoras de diferentes distúrbios neuropsiquiátricos. A construção do jogo foi apoiada em um processo de desenvolvimento composto por várias etapas: estudos teóricos sobre as áreas envolvidas, estudo de tecnologias capazes de apoiar essa integração, levantamento de requisitos com especialistas, implementação e avaliação com especialistas. O produto final foi avaliado por especialistas da área médica, que consideraram os resultados como positivos.

Sim-Colmeia: ambiente de simulação da dinâmica de uma colmeia para o ensino de Biologia

As ferramentas computacionais estão apoiando, de maneira crescente, o processo de ensino e aprendizagem em diversas áreas. Elas aumentam as possibilidades do docente para ministrar um conteúdo e interagir com seus alunos. Neste grupo de ferramentas estão as simulações baseadas em sistemas multiagentes. Neste contexto, este trabalho tem como objetivo apresentar um ambiente de simulação do crescimento populacional de uma colmeia para o ensino de Biologia. As variáveis do sistema podem ser alteradas visando analisar os diferentes resultados obtidos. Aspectos como duração e tempo da florada das plantações, conhecidos como campos de flores, podem ser manipulados pelo aluno. A abordagem multiagentes em Inteligência Artificial Distribuída foi a solução escolhida, para que o controle das atividades do aplicativo fosse feito de maneira automatizada. A Realidade Virtual foi utilizada para acrescentar aspectos importantes do processo que não podem ser visualizados pela simulação matemática. Uma síntese da utilização de tecnologias na educação, em especial da Informática, é discutida no trabalho. Aspectos da aplicação no ensino de Biologia são apresentados, assim como resultados iniciais de sua utilização.