Environments in multiagent systems

Weyns, Danny; Schumacher, Michael; Ricci, Alessandro; Viroli, Mirko; Holvoet, Tom

doi:10.1017/s0269888905000457

Cited by 46 publications

(24 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several researchers have acknowledged that the environments used during the evaluation is an important element to consider [139,138], since such environments establish the rules that agents must follow and represent the abilities that are taken into consideration. Moreover, it is not always clear what benchmarks and testbeds are actually evaluating and how to interpret the results that agents obtain on them [35].…”

Section: Work Focussed On the Interaction Between Agentsmentioning

confidence: 99%

Towards a Universal Test of Social Intelligence

Cabrera¹

View full text Add to dashboard Cite

Under the view of artificial intelligence, an intelligent agent is an autonomous entity which interacts in an environment through observations and actions, trying to achieve one or more goals with the aid of several signals called rewards. The creation of intelligent agents is proliferating during the last decades, and the evaluation of their intelligence is a fundamental issue for their understanding, construction and improvement.Social intelligence is recently obtaining special attention in the creation of intelligent agents due to the current view of human intelligence as highly social. Social intelligence in natural and artificial systems is usually measured by the evaluation of associated traits or tasks that are deemed to represent some facets of social behaviour. The amalgamation of these traits or tasks is then used to configure an operative notion of social intelligence. However, this operative notion does not truly represent what social intelligence is and a definition following this principle will not be precise. Instead, in this thesis we investigate the evaluation of social intelligence in a more formal and general way, by actually considering the evaluee's interaction with other agents.In this thesis we analyse the implications of evaluating social intelligence using a test that evaluates general intelligence. For this purpose, we include other agents into an initially singleagent environment to figure out the issues that appear when evaluating an agent in the context of other agents. From this analysis we obtain useful information for the evaluation of social intelligence.From the lessons learned, we identify the components that should be considered in order to measure social intelligence, and we provide a formal and parametrised definition of social intelligence. This definition calculates an agent's social intelligence as its expected performance in a set of environments with a set of other agents arranged in teams and participating in lineups, with rewards being re-understood appropriately. This is conceived as a tool to define social intelligence testbeds where we can generate several degrees of competitive and cooperative behaviours. We test this definition by experimentally analysing the influence of teams and agent line-ups for several multi-agent systems with variants of Q-learning agents.However, not all testbeds are appropriate for the evaluation of social intelligence. To facilitate the analysis of a social intelligence testbed, we provide some formal property models about social intelligence in order to characterise the testbed and thus assess its suitability. Finally, we use the presented properties to characterise some social games and multi-agent environments, we make a comparison between them and discuss their strengths and weaknesses in order to evaluate social intelligence.ii ResumenBajo la visión de la inteligencia artificial, un agente inteligente es una entidad autónoma la cual interactúa en un entorno a través de observaciones y acciones, tratando de lograr uno o más ...

show abstract

Section: Work Focussed On the Interaction Between Agentsmentioning

confidence: 99%

Towards a Universal Test of Social Intelligence

Cabrera¹

View full text Add to dashboard Cite

show abstract

“…The mini-Swarm system consists of a society of N agents, which cooperate in a socially sharing environment (E) [52] for realizing the common intention of finding high-quality solution for an optimization task.…”

Section: The Mini-swarm Systemmentioning

confidence: 99%

A Mini-Swarm for the Quadratic Knapsack Problem

Xie

Liu

2007

2007 IEEE Swarm Intelligence Symposium

View full text Add to dashboard Cite

Abstract-The 0-1 quadratic knapsack problem (QKP) is a hard computational problem, which is a generalization of the knapsack problem (KP). In this paper, a mini-Swarm system is presented. Each agent, realized with minor declarative knowledge and simple behavioral rules, searches on a structural landscape of the problem through the guided generate-and-test behavior under the law of socially biased individual learning, and cooperates with others by indirect interactions. The formal decomposition of behaviors allows understanding and reusing elemental operators, while utilizes the heuristic information on the landscape. The results on a collection of the QKP instances by mini-Swarm versions are compared with that of both a branch-and-bound algorithm and a greedy genetic algorithm, which show its effectiveness. I. INTRO DUCTIONThe 0-1 quadratic knapsack problem (QKP) was introduced by Gallo et al. [20], which consists in choosing elements from n items for maximizing a quadratic profit objective function subject to a linear capacity constraint. The QKP is a generalization of the 0-1 knapsack problem (KP) [38], by restricting all the quadratic coefficients to zero. The knapsack problems [38] have been intensively studied due to both its theoretical interest and its wide practical applicability. Due to its generality, the QKP has more practically applications [3,20], moreover, several graph problems can be formulated as the QKP [4,45].The KP is NP-hard, although it can be solved exactly in pseudo-polynomial time through dynamic programming based on Bellman recursion [21]. As a generalized version, the QKP is much harder than the KP [10,45], where the graph structure associated with coefficients of objective function plays a important role in its solution [46].Due to the NP-hardness in strong sense, the research efforts on solving the QKP may focus on exact methods, approximate algorithms and heuristic methods.Classical The studies on approximate algorithms, which seek guaranteed polynomial-time performance on approximate solution, were only limited on some special cases of the QKP, such as the classical KP [30], and the QKP where the underlying graph is edge series-parallel [46], etc.The heuristic methods arise from practical interest, which try to find good solution in high probability with acceptable computational efforts by considering certain imprecise knowledge on the structure of the problem. The linearization and exchange (LEX) heuristic [24] uses the best linear L2-approximation to build an associated linear KP and use it for determining items in a greedy way to form a good initial solution, which is then improved by an exchange method between certain items. Hua et al. [29] studied on the convex QKP by an approximate dynamic programming approach. Julstrom [32] also illustrated the capability of several heuristics, including greedy, genetic and greedy genetic algorithms on the QKP.Swarm algorithms [7,16,34] address on tackling a specified optimization task by multiple interacting agents, where each agent works u...

show abstract

“…Their analysis shows three main common dimensions related to (i) the way they are syncing between physical and virtual worlds, (ii) the agents (artificial or human) populating the worlds and cooperating/competing with each other to fulfill their goals, and (iii) the social relations and interactions taking place among the huge number of agents playing different roles in those interactions Ten years ago, Weyns et al formulated in [22] three categories of research challenges as a consequence of following the idea of thinking the environment of a Multiagent System as a first-class citizen: the first category relates to a proper formalization of "environment" making fully clear what is an agent and which element of an overall system is part of the environment. Based on this formal understanding, [22] expected the development of a classification of different kinds of environments also in relation to a corresponding taxonomy of application domains. The second category of challenges drives this further and deals with challenges in exploring the relation between agents and their environment.…”

Section: Introductionmentioning

confidence: 99%

From Physical to Virtual: Widening the Perspective on Multi-Agent Environments

Carrascosa

Klügl

Ricci

et al. 2015

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract. Since more than a decade, the environment is seen as a key element when analyzing, developing or deploying Multi-Agent Systems (MAS) applications. Especially, for the development of multi-agent platforms it has become a key concept, similarly to many application in the area of location-based, distributed systems. An emerging, prominent application area for MAS is related to Virtual Environments. The underlying technology has evolved in a way, that these applications have grown out of science fiction novels till research papers and even real applications. Even more, current technologies enable MAS to be key components of such virtual environments. In this paper, we widen the concept of the environment of a MAS to encompass new and mixed physical, virtual, simulated, etc. forms of environments. We analyze currently most interesting application domains based on three dimensions: the way different "realities" are mixed via the environment, the underlying natures of agents, the possible forms and sophistication of interactions. In addition to this characterization, we discuss how this widened concept of possible environments influences the support it can give for developing applications in the respective domains.

show abstract

Environments in multiagent systems

Cited by 46 publications

References 42 publications

Towards a Universal Test of Social Intelligence

Towards a Universal Test of Social Intelligence

A Mini-Swarm for the Quadratic Knapsack Problem

From Physical to Virtual: Widening the Perspective on Multi-Agent Environments

Contact Info

Product

Resources

About