Calibration de simulations multi-agents à l'aide d'une méthode semi-automatique d'analyse du comportement (présentation courte)

L'un des principaux problemes en simulation multi-agent est la definition des parametres du modele agent et leur calibration. Ce probleme est encore plus difficile lorsqu'on considere des environnements virtuels immersifs, dans les-quels les agents intelligents doivent reproduire des comportements humains et apparaitre « rea-listes » aux yeux des utilisateurs. Dans cet article , nous proposons d'enregistrer et d'analy-ser les comportements des agents pour evaluer leur similarite avec ceux des humains dans un environnement virtuel immersif. Nous utilisons des methodes de classification pour construire une abstraction des comportements individuels. Les types de comportement sont etudies selon la composition de ces classes pour determiner les manques, capacites et erreurs dans le modele agent. Cette methode nous permet 1) d'ecar-ter les jeux de parametres invalides, 2) de ca-librer des simulations valides et 3) d'expliquer les manques du modele agent pour l'ameliorer. Abstract In the context of agent-based simulation, a major issue is to define relevant parameters of the agent model and calibrate them. This issue is yet harder in immersive virtual environments, where intelligent agents reproduce human behaviour and interact with users. In this paper, we propose to log and analyse agents behaviour to evaluate their similarity to humans behaviour in an immersive virtual environment. Clustering is then used to get an abstraction of individual behaviours. The behaviour archetypes are studied in terms of cluster members in order to identify agent lacks, capacities and errors. This study enables to 1) dismiss invalid parameter sets, 2) calibrate valid simulations and 3) explain lacks in the agent models for further improvement.

[1]  Philippe Caillou,et al.  SimAnalyzer: automated description of groups dynamics in agent-based simulations , 2012, AAMAS.

[2]  Nicolas Sabouret,et al.  Analyse des comportements agents par agrégation aux comportements humains , 2014, JFSMA.

[3]  Nigel Gilbert,et al.  MANTA: new experimental results on the emergence of (artificial) ant societies , 2006 .

[4]  Gildas Morvan,et al.  Evidential calibration process of multi-agent based system: An application to forensic entomology , 2012, Expert Syst. Appl..

[5]  S Stradling,et al.  Errors and violations on the roads: a real distinction? , 1990, Ergonomics.

[6]  Patrick Taillandier,et al.  Supervised Feature Evaluation by Consistency Analysis: Application to Measure Sets Used to Characterise Geographic Objects , 2010, 2010 Second International Conference on Knowledge and Systems Engineering.

[7]  Etienne de Sevin,et al.  An evaluation of the COR-E computational model for affective behaviors , 2013, AAMAS.

[9]  Catherine Pelachaud,et al.  Modelling multimodal expression of emotion in a virtual agent , 2009, Philosophical Transactions of the Royal Society B: Biological Sciences.

[10]  N. Gilbert,et al.  Artificial Societies: The Computer Simulation of Social Life , 1995 .

[11]  Alexis Drogoul,et al.  MANTA: New Experimental Results on the Emergence of (Artificial) Ant Societies , 1995 .

[12]  Mohamed Chetouani,et al.  Interpersonal Synchrony: A Survey of Evaluation Methods across Disciplines , 2012, IEEE Transactions on Affective Computing.

[13]  Frank Puppe,et al.  Approaches for resolving the dilemma between model structure refinement and parameter calibration in agent-based simulations , 2006, AAMAS '06.

[14]  Benoit Lacroix,et al.  Formalizing the construction of populations in multi-agent simulations , 2013, Eng. Appl. Artif. Intell..

[15]  James C. Lester,et al.  The persona effect: affective impact of animated pedagogical agents , 1997, CHI.

[16]  Philippe Mathieu,et al.  A Generic Architecture for Realistic Simulations of Complex Financial Dynamics , 2010, PAAMS.

[17]  Sylvain Piechowiak,et al.  A behavioral multi-agent model for road traffic simulation , 2008, Eng. Appl. Artif. Intell..

[18]  Michel C. A. Klein,et al.  Agent-Based Analysis of Patterns in Crowd Behaviour Involving Contagion of Mental States , 2011, IEA/AIE.

[19]  Nicolas Sabouret,et al.  Agents Behavior Semi-automatic Analysis through Their Comparison to Human Behavior Clustering , 2014, IVA.

[20]  Frank Puppe,et al.  Techniques for Analysis and Calibration of Multi-agent Simulations , 2004, ESAW.