LEMMING: A learning system for multi-robot environments

Abstract

Describes LEMMING, a learning system for multiple mobile robot environments. LEMMING extends the idea of the broadcast-based contract net protocol with CBR (case-based reasoning). In terms of the CBR mechanism, LEMMING can learn to select appropriate robots for a given task. Thus, LEMMING makes it possible to reduce the waste of communication resources and the trouble of processing irrelevant messages. The paper evaluates LEMMING with some experiments and shows that LEMMING can deal with task negotiation more effectively than a broadcast-based contract net system.