Learning social behaviors without sensing

  • Authors:

  • Anand Panangadan;Michael G. Dyer

  • Affiliations:

  • Computer Science Department, University of California, Los Angeles, Los Angeles, California;Computer Science Department, University of California, Los Angeles, Los Angeles, California

  • Venue:
  • ICSAB Proceedings of the seventh international conference on simulation of adaptive behavior on From animals to animats
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

A learning algorithm is presented that enables agents that do not have the ability to sense other agents to adapt its behaviors (that were learned in a single agent environment) to novel situations (deadlocks arising from existing in an autonomous multi-agent system). This adaptation takes place as the agent continues to perform its construction task. When the agents are confined to narrow spaces, this learned behavior leads to a "bucket brigade". The algorithm also learns the pattern of activations on its spatial map that is associated with deadlocks and the new behaviors are exhibited when this pattern is later observed.