Organizations are an important basis for coordination in multiagent systems. However, there is no best way to organize and all ways of organizing are not equally effective. Attempting to optimize an organizational structure depends strongly on environmental features including problem characteristics, available resources, and agent capabilities. If the environment is dynamic, the environmental conditions or the problem task structure may change over time. This precludes the use of static, design-time generated, organizational structures in such systems. On the other hand, for many real environments, the problems are not totally unique either: certain characteristics and conditions change slowly, if at all, and these can have an important effect in creating stable organizational structures.

Organizational-Self Design (OSD) has been proposed as an approach for constructing suitable organizational structures at runtime. We extend the existing OSD approach to include worthoriented domains, model other resources in addition to only processor resources and build in robustness into the organization. We then evaluate our approach against the contract-net approach and show that our OSD agents perform better, are more efficient, and more flexible to changes in the environment.