Skip to main content

Über dieses Buch

This book constitutes the thoroughly refereed post-conference proceedings of the 15th International Workshop on Multi-Agent-Based Simulation, MABS 2014, held in Paris, France, in May 2014. The workshop was held in conjunction with the 13th International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2014.

The 17 revised full papers included in this volume were carefully selected from numerous submissions. The papers are organized in topical sections on simulation methodologies, simulation of social behaviour, data and multi-agent-based simulation and applications.



Simulation Methodologies


Event-Driven Multi-agent Simulation

Most agent-based models today apply a time-driven approach, i.e. simulation time is advanced in equidistant steps. This time advance method is considerably easier to implement than the more flexible and efficient event-driven approach.
Applying the event-driven approach requires that (a) the durations for agent and environment actions are determined before they terminate, (b) each agent is able to instantly react to changes in its environment, and (c) the update of the state of the environment can be kept efficient despite updating agents asynchronously.
The simulation toolkit famos fulfils these requirements, extending an existing discrete-event simulator. The toolkit also supports a flexible representation of space and the movement of agents in that space. These are areas where existing toolkits for agent-based modelling show shortcomings, despite the fact that a majority of multi-agent models explicitly model space and allow for mobile agents.
Ruth Meyer

RatKit: Repeatable Automated Testing Toolkit for Agent-Based Modeling and Simulation

Agent-based modeling and simulation (ABMS) became an attractive and efficient way to model large-scale complex systems. The use of models always raises the question whether the model is correctly encoded (verification) and accurately represents the real system (validation). However, achieving a sufficiently credible agent-based simulation (ABS) model is still difficult due to weak verification, validation and testing (VV&T) techniques. Moreover, there is no comprehensive and integrated toolkit for VV&T of ABS models that demonstrates that inaccuracies exist and/or which reveals the existing errors in the model. Based on this observation, we designed and developed RatKit: a toolkit for ABS models to conduct VV&T. RatKit facilitates the VV&T process of ABMS by providing an integrated environment that allows repeatable and automated execution of ABS tests. This paper presents RatKit in detail and demonstrates its effectiveness by showing its applicability on a simple well-known case study: predator - prey.
İbrahim Çakırlar, Önder Gürcan, Oğuz Dikenelli, Şebnem Bora

Man on Earth – The Challenge of Discovering Viable Ecological Survival Strategies

Many previous societies have killed themselves off and, in the process, devastated their environments. Perhaps the most famous of these is that of “Easter Island”. This suggests a grand challenge: that of helping discover what kinds of rationality and/or coordination mechanisms might allow humans and the greatest possible variety of other species to coexist. As their contribution towards this, the agent community could investigate these questions within simulations to suggest hypotheses as to how this could be done. The particular problem for our community is that of designing and releasing a society of plausible agents into a simulated ecology and assessing: (a) whether the agents survive and (b) if they do survive, what impact they have upon the diversity of other species in the simulation. No other community is currently in a position to explore this problem as a whole. The simulated ecology needs to implement a suitably dynamic, complex and reactive environment for the test to be meaningful. In such a simulation, agents (as any other entity) would have to eat other entities to survive, but if they destroy the species they depend upon they are likely to die off themselves. Up to now there has been a lack of simulations that combine a complex model of the ecology with a multi-agent model of society – there have been complex models of society but with simple ecological representations and complex ecological models but with little of human social complexity in them. In order for progress to be made with humanity’s challenge, we will have to move beyond simple ideas and solutions and embrace the complexity of the socio-ecological complex as a whole. A suitable dynamic ecological model and simple tests with agents are described to illustrate this challenge, as the first steps towards a meaningful test bed to under pin the implied research programme.
Bruce Edmonds

Modelling Environments in ABMS: A System Dynamics Approach

Environment is a basic concept of agent-based modeling and simulation (ABMS), in which agents exist and interact, can perceive and act. Common approaches to model environment in ABMS consider environment as a physical space (typically a 2D grid) or as a virtual space that supports agent to agent interaction. In this paper, we introduce a method for modeling the environment in agent-based simulations that integrates both its physical and social aspects. Borrowing from System Dynamics, we specify environment as a set of stocks and flows. Stocks represent physical resources, spatial locations and social structures in a uniform way. Stocks can be perceived and modified by agents through flows. Our method considers and describes the environment as a combination of spatial space (or network), physical structure, and social structure, which makes the inter-relation between the global state environment variables and the behavior of the agents explicit. We illustrate the use of the method through its application to a case study in consumer lighting.
Reza Hesan, Amineh Ghorbani, Virginia Dignum

Simulation of Social Behaviour


Modeling Culturally-Influenced Decisions

This article proposes a model of culturally influenced decision processes. In particular, cultures influence individual motivation, jointly with human nature and personality. The use of this model is then illustrated by a simulation model of the impact of cultural differences on organizational performance (efficiency, flexibility and member satisfaction) in two organizational structures (bureaucracies and adhocracies). This model is validated against empirical evidence from social sciences.
Loïs Vanhée, Frank Dignum, Jacques Ferber

Gender Differences: The Role of Nature, Nurture, Social Identity and Self-organization

This paper describes an agent-based model to investigate the origins of gender differences in social status. The agents’ basic behaviour is modelled according to Kemper’s sociological status-power theory. Differences in the socializing forces of the surrounding society are modelled using Hofstede’s dimensions of culture. Particulars of play behaviour are modelled using experimental child development studies from various cultures. The resulting model is presented and discussed. Social identity as a group of either non-gendered children, boys, or girls, seems a powerful force, multiplying the effect of biological differences. The model is actually general enough to be applicable to a wide range of social behaviours with minimal changes.
Gert Jan Hofstede, Frank Dignum, Rui Prada, Jillian Student, Loïs Vanhée

Partner Selection Delays Extinction in Cooperative and Coordination Dilemmas

Multiagent systems have been used to model and study social systems. Such studies have focused on cooperation and coordination dilemmas. The goal was to investigate how a population of agents could escape those dilemmas. Typically those studies assume large populations either fixed size or infinite. However, when we introduce variable sized population, a new risk arises consisting on population extinction, which is a stable point of the corresponding dynamics. We present the Energy Based Evolutionary Algorithm, a model where agents are born, interact, reproduce and die. Interaction is mediated by some game which is the sole means of acquiring energy needed for reproduction. In this paper we show that when an agent is capable of selecting its partners based on knowledge of successful interactions, the population is able to survive longer when compared with random partner selection. We present results using a set of well known games.
Pedro Mariano, Luís Correia

Group Size and Gossip Strategies: An ABM Tool for Investigating Reputation-Based Cooperation

In an environment in which free-riders are better off than cooperators, social control is required to foster and maintain cooperation. There are two main paths through which social control can be applied: punishment and reputation. Using a Public Goods Game, we show that gossip, used for assortment under three different strategies, can be effective in large groups, whereas its efficacy is reduced in small groups, with no main effect of the gossiping strategy. We also test four different combinations of gossip and costly punishment, showing that a combination of punishment and reputation-based partner selection leads to higher cooperation rates.
Francesca Giardini, Mario Paolucci, Diana Adamatti, Rosaria Conte

Data and Multi-agent-Based Simulation


Automatic Generation of Agent Behavior Models from Raw Observational Data

Agent-based modeling is used to simulate human behaviors in different fields. The process of building believable models of human behavior requires that domain experts and Artificial Intelligence experts work closely together to build custom models for each domain, which requires significant effort. The aim of this study is to automate at least some parts of this process. We present an algorithm called , which produces an agent behavioral model from raw observational data. It calculates transition probabilities between actions and identifies decision points at which the agent requires additional information in order to choose the appropriate action. Our experiments using synthetically-generated data and real-world data from a hospital setting show that the algorithm can automatically produce an agent decision process. The agent’s underlying behavior can then be modified by domain experts, thus reducing the complexity of producing believable agent behavior from field data.
Bridgette Parsons, José M. Vidal, Nathan Huynh, Rita Snyder

Data Analysis of Social Simulations Outputs - Interpreting the Dispersion of Variables

In the domain of social simulation, there are very few papers reporting on the statistical analysis of simulation results, while it is very common in empirical social sciences. The paper advocates the recourse to the statistical analysis of social simulation outputs, as a very efficient way to improve the interpretation of simulation results and so the understanding of the system that is the model’s target. This is illustrated by the study of a simulation model designed to analyze a real case regarding the management of a river in South West of France. Several standard statistics methods are used to shed light on the possible outcomes of the debate between the stakeholders.
Christophe Sibertin-Blanc, Nathalie Villa-Vialaneix

Reproducing and Exploring Past Events Using Agent-Based Geo-Historical Models

The field of “digital humanities” is about using the latest digital methodologies in order to tackle humanities disciplines and social sciences questions. The ARCHIVES project belongs to this new research area. It proposes a methodology to build agent-based models of historical events, in particular crisis events, in order to answer new questions about them or explore them in new ways. In this paper, we present the first implementation of ARCHIVES on the case study of the management of floods in Hà Nội (Việt Nam) in 1926. We show how we collected, digitized and indexed numerous historical documents from various sources, built a historical geographic information system to represent the environment and flooding events and finally designed an agent-based model of human activities in this reconstructed environment. We then show how this model helped us understanding the decisions made by the different actors during this event, testing multiple scenarios and answering several questions concerning the management of the flooding events.
Nasser Gasmi, Arnaud Grignard, Alexis Drogoul, Benoit Gaudou, Patrick Taillandier, Olivier Tessier, Vo Duc An



TENDENKO: Agent-Based Evacuation Drill and Emergency Planning System

Evacuation drills are conducted periodically to practice smooth evacuations from buildings and rescue operations at emergency sites. An agent-based evacuation simulation provides a platform for simulating human evacuation behavior during emergencies, which can be affected by various social and human factors. These factors include agent characteristics, societal behavior codes, evacuation guidance, and so on. These factors make it difficult to conduct evacuation drills and develop prevention plans for unexpected emergencies. TENDENKO aims to simulate evacuation drills at buildings where real drills cannot be conducted, and to improve evacuation planning for the building to save more lives during future emergencies.
Toshinori Niwa, Masaru Okaya, Tomoichi Takahashi

Analysing the Apprenticeship System in the Maghribi Traders Coalition

In this work we further the investigation into the functioning of the Maghribi Traders Coalition – a historically significant traders collective that operated along the North African coast between the 10th and 13th centuries. They acted as a closed group whose interactions were governed by informal institutions (i.e. norms). Historical accounts point to an apprenticeship system that was in force in this society. In this work we propose an agent-based model of the society with the apprenticeship mechanism and analyse the role the mechanism may have played in the removal of cheaters from their trade relationship networks.
Christopher Frantz, Martin K. Purvis, Mariusz Nowostawski, Bastin Tony Roy Savarimuthu

Spatial Modeling of Agent-Based Prediction Markets: Role of Individuals

In this paper, we extend the spatial agent-based prediction market proposed by Yu and Chen at MABS 2011 into a spatial model in which agents choose their community (neighbors) by following Schelling’s proximity model. This extended model generalizes the spatial configuration of the original model and enables us to examine the validity of the Hayek hypothesis when the information distribution is determined by clusters of agents with heterogeneous identities. Specifically, we examine the role of the toleration capacity, the key parameter in the Schelling model, which generates the clusters of agents with different sizes, and the role of exploration capacity which determines how well an agent is informed about his local surroundings. We find that after taking into account market activity and price volatility, both the toleration capacity and exploration capacity have a positive effect on the prediction accuracy and enhance information polling and the information aggregation of markets. The results obtained in this agent-based simulation, therefore, add a qualification to the well-known Hayek hypothesis and point to the significance of individuals in information aggregation.
Bin-Tzong Chie, Shu-Heng Chen

Agent Based Exploration of Urban Economic Dynamics Under the Rent-Gap Hypotheses

We present a stylised agent-based model of housing investment based on the rent gap theory proposed by the late Neil Smith. We couple Smith’s supply-side approach to investment, with individual-level residential mobility within a city. The model explores the impact of varying levels of capital flowing in the city and reproduces certain theorised and observed dynamics emerging from the cyclic nature of investment: the tendency of capital to spatially concentrate generating intra-urban inequalities, the occasional formation of persistent pockets of disinvestment and phenomena such as gentrification.
Stefano Picascia, Bruce Edmonds, Alison Heppenstall

Emergent Collective Behaviors in a Multi-agent Reinforcement Learning Pedestrian Simulation: A Case Study

In this work, a Multi-agent Reinforcement Learning framework is used to generate simulations of virtual pedestrians groups. The aim is to study the influence of two different learning approaches in the quality of generated simulations. The case of study consists on the simulation of the crossing of two groups of embodied virtual agents inside a narrow corridor. This scenario is a classic experiment inside the pedestrian modeling area, because a collective behavior, specifically the lanes formation, emerges with real pedestrians. The paper studies the influence of different learning algorithms, function approximation approaches, and knowledge transfer mechanisms on performance of learned pedestrian behaviors. Specifically, two different RL-based schemas are analyzed. The first one, Iterative Vector Quantization with Q-Learning (ITVQQL), improves iteratively a state-space generalizer based on vector quantization. The second scheme, named TS, uses tile coding as the generalization method with the Sarsa(\(\lambda \)) algorithm. Knowledge transfer approach is based on the use of Probabilistic Policy Reuse to incorporate previously acquired knowledge in current learning processes; additionally, value function transfer is also used in the ITVQQL schema to transfer the value function between consecutive iterations. Results demonstrate empirically that our RL framework generates individual behaviors capable of emerging the expected collective behavior as occurred in real pedestrians. This collective behavior appears independently of the learning algorithm and the generalization method used, but depends extremely on whether knowledge transfer was applied or not. In addition, the use of transfer techniques has a remarkable influence in the final performance (measured in number of times that the task was solved) of the learned behaviors.
Francisco Martinez-Gil, Miguel Lozano, Fernando Fernández

Cognitive Modeling of Behavioral Experiments in Network Science Using ACT-R Architecture

The Network Science has dedicated a considerable amount of effort to the study of many distributed collective decision-making processes which must balance diverse individual preferences with an expectation for collective unity. Several works have reported their results about behavioral experiments on biased voting in networks individuals, however we will focus on the results reported on [1] on which were run 81 experiments, on which participated 36 human subjects arranged in a virtual network who were financially motivated in a heterogeneous manner and whose goal was to reach global consensus to one of two opposing choices. Multiple experiments were performed using diverse topological network configurations, different schemes of financial incentives that created opposing tensions between personal preferences, and finally different ratios of both inter and intra-connectivity among the network nodes. The corresponding analysis of the results demonstrated that changing those features of the experiments produced different kind of social behavioral patterns as a result. Thus, the purpose of this work is manifold: on the one hand, it aims to describe the possible structures that underlie the decision-making process of these experiments through the modeling of symbolic cognitive prototypes supported by a robust and complex cognitive architecture so-called ACT-R and, on the other hand, by applying modifications in the ACT-R parameters to find those subtle aspects that can either influence both the performance and speed of convergence of the experiments or cause the total inability to reach a global consensus in a reasonable amount of time.
Oscar J. Romero, Christian Lebiere


Weitere Informationen