Heuristiscs-Based High-Level Strategy for Multi-agent Systems

  • Authors:

  • Péter Gasztonyi;István Harmati

  • Affiliations:

  • Department of Control Engineering and Information Technology, Budapest University of Technology and Economics, Budapest, Hungary H-2117;Department of Control Engineering and Information Technology, Budapest University of Technology and Economics, Budapest, Hungary H-2117

  • Venue:
  • ICANN '08 Proceedings of the 18th international conference on Artificial Neural Networks, Part I
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, a high-level strategy concept is presented for robot soccer, based on low level heuristic inference methods, rather than explicit rule-based strategy. During tactical positioning, no strict role set is assigned for the agents, instead a fitting point of the role-space is selected dynamically. The algorithm for this approach applies fuzzy logic. We compute fields-of-quality, regarding some relevant aspects of the scenario, and integrate them into one decision-field, according to given strategic parameters (used as weights). The most relevant locations are derived from the decision-field through subtractive clustering, and the agents are allocated to these locations, as their desired positions, according to their significance and their cost of reaching the given target. If an agent is in a position to manipulate the ball, an appropriate action is being selected for it. The simulation and experiments prove that the proposed approach can be efficient in dynamically changing environment or against opponents of different strategies.