Cédric Sanza

A Distributed Virtual Storytelling System for Firefighters Training

The aim of the proposed paper is to present a virtual reality scenario managing application which has been distributed according to the DoD (Department of Defense) High Level Architecture (HLA) standard. The system works over several implementations of the Run-Time Infrastructure (RTI, HLA software part), on networked PCs. The main scenario is a networked Firefighters Training Simulation where each computer can control one or several avatars either through human piloting or through Artificial Intelligence (AI) based behaviors using Learning Classifier Systems.

Evolving prediction weights using evolution strategy

The evolution strategy is one of the strongest evolutionary algorithms for optimizing real-value vectors. In this paper, we study how to use it for the evolution of prediction weights in XCSF in order to make the computed prediction more accurate. Our version of XCSF shows to be able to evolve more accurate linear approximations of functions. It is more efficient than the original XCSF and slightly better than XCSF with recursive least squares, in spite of its simple structure and its low complexity.

Evolution and cooperation of virtual entities with classifier systems

This paper presents a behavioral system based on artificial life paradigms. The system, called αCS. is suited to be employed for the animation of virtual entities immersed in concurrent and changing environments. The αCS system is the extension of an original classifier system to collaborative abilities. The main modifications enable αCS to use cooperation and communication to build dynamically the behavior of virtual entities which goal is to achieve several tasks. Our classifier system is evaluated by the way of two applications. Firstly, we present the performances of αCS in an optimization problem consisting on following a moving target. Secondly, we investigate a complex 3D world where autonomous entities and avatars interact. Through the simulation of a virtual game, we show how the integration of our system in virtual entities enables to build evolving behaviors thanks to adaptation, communication and auto-organization.

Communication and Interaction with Learning Agents in Virtual Soccer

This paper presents a learning system based on Artificial Life for the animation of virtual entities. The model uses an extension of a classifiers system to build dynamically the behavior of agents by emergence. A behavior is selected into a set of binary rules that evolves continuously to ensure the maximization of predefined goals. The reinforcement allows to reward a rule and then to evaluate its efficiency faced to a given context. We investigate the interaction between virtual agents and a human controlled clone immersed in virtual soccer. In the simulation, each entity evolves in real-time by using the ability of cooperation and communication with teammates. We evaluate the benefits of the communication inside a team and present how it can improve the learning of a group thanks to a rule-sharing and a human intervention.

Constrained control of non-playing characters using Monte Carlo Tree Search

In this paper, we apply the Monte Carlo Tree Search (MCTS) method for controlling at once several virtual characters in a 3D multi-player learning game. The MCTS is used as a search algorithm to explore a search space where every potential solution reflects a specific state of the game environment. Functions representing the interaction abilities of each character are provided to the algorithm to leap from one state to another. We show that the MCTS algorithm successfully manages to plan the actions for several virtual characters in a synchronized fashion, from the initial state to one or more desirable end states. Besides, we demonstrate the ability of this algorithm to fulfill a specific requirement of a learning game AI : guiding the non player characters to follow a predefined and constrained learning scenario and, if necessary, to adapt their decision to unexpected events in the simulation.

Intuitive Method for Pedestrians in Virtual Environments

Recent works about pedestrian simulation can actually be sorted in two categories. The first ones focusing on large crowd simulation aim to solve performance and scalability issues at the expense of behavioral realism of each simulated individual. The second ones aim at individual behavioral realism but the computational cost is too expensive to simulate crowds. In this paper, we propose an alternate approach combining a light reactive behavior with cognitive strategies issued from real life videos. This approach aims at the real time simulation of small crowds of pedestrians (one to two hundred individuals) but with concerns for visual realism regarding heterogeneous behaviors, trajectories and positioning on sidewalks.

Control of non player characters in a medical learning game with Monte Carlo tree search

In this paper, we apply the Monte Carlo Tree Search (MCTS) method for controlling at once several virtual characters in a 3D multi-player learning game. The MCTS is used as a search algorithm to explore a search space where every potential solution reflects a specific state of the game environment. Functions representing the interaction abilities of each character are provided to the algorithm to leap from one state to another. We show that the MCTS algorithm successfully manages to plan the actions for several virtual characters in a synchronized fashion, from the initial state to one or more desirable end states. Besides, we demonstrate the ability of this algorithm to fulfill two specific requirements of a learning game AI : guiding the non player characters to follow a predefined plan while coping with the unpredictability of the human players actions.

Theatrical Text to 3D Virtual Scenography

Generation and visualization systems based on the principle of text-to-scene are powerful systems but they are not dedicated to the particular domain of theatre. Text-to-scene systems provide an intuitive way to easily create 3D environments for non-expert users. In this paper we propose a text-to-scene system dedicated to the theatre, which generates a 3D view based on the descriptive elements extracted from the theatrical text. Such a tool is designed to encourage the emergence of new methods in staging, research and teaching in the domain of the theatre. The works presented in this paper enable the elements of a theatrical production to be manipulated, with both a textual view and a 3D view. The 3D view is obtained from an automatic generation of a 3D environment, based on elements of scenography identified into the text by the user. Semantic annotations are added into the theatrical text to automatically obtain a visualization of a 3D virtual scenography. To encode the descriptive properties, the user adds tags, which identify semantic and descriptive information into the theatrical text. The tags are then used to calculate the location and the orientation of characters and objects in the resulting virtual scenography. The tags are also used to identify the emotional states, the postures of the characters, as well as the duration of each state in order to generate views of the scene.