Stéphane Calderoni

Towards an emergence machine for complex systems simulations

This paper presents a simulation platform that has been realised in Java 1.1 for the study of behaviour and evolutionary processes in non-linear systems. To support the modelling of such systems, we propose the use of agent technology as high level tool to design applications. The framework enables to study emergence by exploiting distributing computing as key issues of the system behaviour. Applications developed with the platform are then simulated to adequately capture any behaviour likely to be observed, to exhibit self organised structures, and to emphasise complex processes, which are brought in action. This approach then allows the study of macroscopic collections endowed with the potential to evolve during time.

MUTANT: a multiagent toolkit for artificial life simulation

Nowadays many artificial life research areas resort to agent-based simulation. This fact has brought us to design a powerful generic platform that would allow scientists in those fields of research to easily build simulation environment. This platform called MUTANT includes a model of self-adaptive agent with genetic evolving capabilities as learning mechanisms. The platform also includes a powerful graphical user interface providing many tools for both modeling and simulation. There are tools for agents design behaviors programming environments description and observation of running simulations. MUTANT is being developed in Java with the aim of being directly usable throughout the Internet.

Genetic Progamming for Automatic Design of Self-Adaptive Robots

The general framework tackled in this paper is the automatic generation of intelligent collective behaviors using genetic programming and reinforcement learning. We define a behavior-based system relying on automatic design process using artificial evolution to synthesize high level behaviors for autonomous agents. Behavioral strategies are described by tree-based structures, and manipulated by genetic evolving processes. Each strategy is dynamically evaluated during simulation, and weighted by an adaptative value. This value is a quality factor that reflects the relevance of a strategy as a good solution for the learning task. It is computed using heterogeneous reinforcement techniques associating immediate and delayed reinforcements as dynamic progress estimators. This work has been tested upon a canonical experimentation framework: the foraging robots problem. Simulations have been conducted and have produced some promising results.

GEAMAS V2.0: an object oriented platform for complex systems simulations

This paper presents the object oriented design and implementation of GEAMAS V2.0, a toolkit for virtual simulations of complex systems. GEAMAS V2.0 is structured in three modules: the kernel, the generation environment and the simulation environment. The kernel implements an object model for agents and provides generic classes. The generation environment allows the graphical design of applications. The simulation environment enables the observation of the simulations evolution via graphical user interface tools. The implementation uses Java 1.1. We applied GEAMAS V2.0 to the sand-pile automaton problem. This reference application, which is easily modeled with GEAMAS V2.0, provides a first validation of our system architecture and simulation mechanism.

JAAFAAR: A Web-Based Multi-Agent Toolkit for Collective Research

This paper introduces a generic and opened multi-agent platform that provides a powerful scientific equipment for collective research on self-organized systems. A general thought on the mutation of object model towards the agent model is presented. The paper details the construction of the platform upon generic models of environment and agent. Then we present the extension of the toolkit to web-based technologies, and its suitability for collective research and remote works.

MUTANT: A Genetic Learning System

This paper presents MUTANT, a learning system for autonomous agents. MUTANT is an adaptive control architecture founded on genetic techniques and reinforcement learning. The system allows an agent to learn some complex tasks without requiring its designer to fully specify how they should be carried out. An agent behavior is defined by a set of rules, genetically encoded. The rules are evolved over time by a genetic algorithm to synthesize some new better rules according to their respective adaptive function, computed by progressive reinforcements. The system is validated through an experimentation in collective robotics.

Genetic encoding of agent behavioral strategy

The general framework tackled in this paper is the automatic generation of intelligent collective behaviors using genetic programming and reinforcement teaming. We define a behavior-based system relying on automatic design process using artificial evolution to synthesize high level behaviors for autonomous agents. Behavioral strategies are described by tree-based structures, and manipulated by generic evolving processes. Each strategy is dynamically evaluated during simulation, and is weighted by an adaptation function as a quality factor that reflects its relevance as good solution for the learning task. It is computed using heterogeneous reinforcement techniques associating immediate reinforcements and delayed reinforcements as dynamic progress estimators.

Zooming on a multiagent simulation system: from the conceptual architecture to the interaction protocol

GEAMAS is a knowledge engineering environment for multi-agent simulation of complex systems. The basic architecture of GEAMAS is designed around three dimensions: MultiAgent Systems (MAS) software design, MAS knowledge abstraction and MAS services dimension. Each one of this three aspects, is implemented into several modular open software layers. First, the paper argues the benefits of a such architecture for a MAS environment. We especially present: 1) how the MAS software design dimension enables to define appropriate tools levels for designing multi-agent systems; 2) how MAS knowledge abstraction adds significant value to implement a computational model of agents; 3) how MAS services dimension allows to correctly extend the GEAMAS environment by integrating new multi-agent concepts such as organization capacities or learning mechanisms.