An experimental benchmarking of two multi-agent architectures for production scheduling and control

The paper provides a first experimental contribution to the validation and comparison of multi-agent heterarchical architectures in order to point out their limits, their main peculiarities and real applicability. For this purpose, the behaviours of two particular control solutions, characterised by a different degree of decision-making delegation assigned to physical and information units in the production system, have been simulated. The experimental campaign is divided into three trials, in order to measure the performances of the two architectures not only under stable and predictable conditions, but also in case of unexpected events, such as the release of urgent manufacturing orders and the occurrence of failures in production resources. (C) 2000 Elsevier Science B.V. All rights reserved.