Simulating robot groups with elements of a social structure using KVORUM

摘要

This paper describes the architecture of KVORUM, the agent-based modeling environment developed by the authors to explore certain design decisions and facilitate simulation usage in the field of group robotics, specifically for research of robotic systems with elements of a social structure. KVORUM is a prototype of a simulation system that should adequately abstract away complexities of a physical system while providing convenient interfaces and library modules for modeling groups of mobile land robots and individual agent's intrinsic structures. It was built from the ground up as a modular, highly extensible system focused on speed of calculations that is to be achieved by using proper simplification of physical and other effects, and by the ability to perform simulations using parallel computing. In KVORUM the developed abstractions and limitations of a parallel architecture are tested without fully adhering to parallel computing requirements. It is shown that the proposed method of simulating such systems is applicable to a wide variety of problems from the field of group robotics with elements of a social structure (and swarm robotics in general). Some features of the architecture and models used in KVORUM allow for it to be extended in next iterations to fully support simulations on parallel computing systems.