Robustness in organizational-learning oriented classifier system

Abstract An organizational-learning oriented classifier system (OCS) is an extension of learning classifier systems (LCSs) to multiagent environments, where the system introduces the concepts of organizational learning (OL) in organization and management science. To investigate the capabilities of OCS as a new multiagent-based LCS architecture, this paper specifically focuses on the robustness of OCS in multiagent environments and explores its capability in space shuttle crew task scheduling as one of real-world applications. Intensive simulations on a complex domain problem revealed that OCS has robustness capability in the given problem. Concretely, we found that OCS derives the following implications on robustness: (1) OCS finds good solutions at small computational costs even after anomaly situations occur; and (2) this advantage becomes stronger as the number of anomalies increases.