The JaCalIVE framework for MAS in IVE: A case study in evolving modular robotics

Abstract

This paper presents a framework specially designed for the execution and adaptation of Intelligent Virtual Environments. This framework, called JaCalIVE, facilitates the development of this kind of environments managing in an efficient and realistic way the evolution of parameters for the adaptation of the physical world. The framework includes a design method and a physical simulator which is in charge of giving the Intelligent Virtual Environment the look of the real or physical world, allowing to simulate physical phenomena such as gravity or collision detection. The paper also includes a case study which illustrates the use of the proposed framework as an evolutive algorithm which allows the automatic adaptation of modular robots.

Introduction

Nowadays, having software solutions at one’s disposal that enforce autonomy, robustness, flexibility and adaptability of the system to develop is completely necessary. The dynamic agents organizations that auto-adjust themselves to obtain advantages from their environment seems a more than suitable technology to cope with the development of this type of systems. These organizations could appear in emergent or dynamic agent societies, such as grid domains, peer-to-peer networks or other contexts where agents dynamically group together to offer compound services as in Intelligent Virtual Environments (IVE). An IVE is a virtual environment simulating a physical (or real) world, inhabited by autonomous intelligent entities [1].

Today, this kind of applications are between the most demanded ones, not only as being the key for multi-user games such as World Of Warcraft¹ (with more than 7 million of users in 2013)² but also for immersive social networks such as Second Life³ (with 36 million accounts created in its 10 years of history)⁴. It is in the development of these huge IVEs where the need of a quick and easy-to-use modeling toolkit arises.

These kinds of IVEs are addressed to a huge number of simultaneous entities, so they must be supported by highly scalable software. This software has also to be able to adapt to changes, not only of the amount of entities but also of their users needs. Technology currently used to develop this kind of products lacks of elements facilitating the adaptation and management of the system. Traditionally, this kind of applications use the client/server paradigm, but due to their features, a distributed approach such as multi-agent systems (MAS) seems to fit in the development of components that will evolve in an autonomous way and coordinated with the own environment’s evolution. In the last decade, MAS technology has been successfully employed in similar large scale distributed systems such as Robocup Rescue simulation [2].

This paper presents the JaCalIVE⁵ (Jason Cartago implemented Intelligent Virtual Environment) framework. It provides a method to develop this kind of IVEs along with a supporting platform to execute them. JaCalIVE is based on the MAM5 meta-model [3] which describes a method to design IVEs.

MAM5 is based in the A&A (Agent & Artifact) meta-model [4] that describes environments for MAS as populated not only by agents, but also for other entities that are called artifacts. The A&A meta-model promotes the modeling and engineering of agent societies and MAS environment as first-class entities. According to MAM5 meta-model, an IVE is composed of three important parts: artifacts, agents and physical simulation. Artifacts are the elements in which the environment is modeled. Agents are the IVE intelligent part. The physical simulation is in charge of giving the IVE the look of the real or physical world, allowing to simulate physical phenomenal such as gravity or collision detection.

In order to evaluate the proposed framework we have chosen a case study consisting in the implementation of an evolutive algorithm which allows the automatic creation of modular robots optimized for a specific task. Specifically, the implemented system simulates a genetic algorithm where robots can interact among them in order to change its shape by joining other modules or environment objects. In that sense, one simple modular robot can change its shape and create a complex robot depending on the movement required. The aim of each robot is to minimize the distance between its real movement and the movement defined by the fitness function. During the simulation robots will evolve or will be destroyed following the rules of an evolutive algorithm. The modular robots obtained at the end of the simulation will be the better adapted and the most appropriated to do the required movement.

The rest of the paper is organized as follows: Section 2 summarizes the most important related work. Section 3 describes the JaCalIVE framework. Section 4 presents the proposed case study based on the automatic evolution of modular robotics developed using JaCalIVE. Finally, Section 5 summarizes the main conclusions of this work.