I. Nagy and A.L. Bencsik (Hungary)
Behaviour-oriented, Evolutionary, Genetic algorithm, Mobile Robot, Multi-agent, Potential Field, Self Organized.
This paper is describing a potential field building simulation with mobile robots on multi-agent domains. The paper contains description of the structure and operation of the simulation system, and, at the end, through an example a report on the results. In the example a group of mobile robots, equipped with ultrasonic range measuring sensors, is building up the potential field of the work space. The system, on the one hand, offers some solutions for the problems related to the potential field building (like dead lock problem), and on the other hand, looks for the problem-solutions connected with self organization of the mobile robots. The system itself can be characterized as a decentralized, behaviour-oriented, and evolutionary self-organized simulation system. It is decentralized and behaviour-oriented, because the agents sharing the basic information about the position and orientation between each other, and on the basis of these information they define the next possible position and orientation. It is evolutionary self-organized, because the moving strategies are defined by a genetic algorithm, through which and the specified policies the near-optimal next possible move can be determined. The whole simulation system has been prepared in MATLAB 5.1 software environment. The structure of the paper – thematically it can be divided into three main parts. The first part describes the ultrasonic (US) sensing and range measuring with the systematic errors deriving from this measuring, the potential field building, and the classical moving strategies, first on single- then on multi-agent domains. Afterwards, a simple moving-mechanism and behaviour based mechanism is described. The second part of the paper contains the description of the genetic algorithm (GA). At the end, in the appendixes – 3rd part –, the obtained results are revealed.
Important Links:
Go Back
