Currently, large systems are described in terms of entities that act as providers and consumers. These entities offer their functionality through services and interact to provide or ask for these services. The integration of Open Multi-Agent Systems and Service-Oriented Computing seems to be suitable to implement these systems. In open MAS, agents enter and leave the system, interact with others in a flexible way, and are considered as reactive and proactive entities that are able to reason about what is happening in their environment and perform local actions based on their observations to achieve their goals. Service-Oriented Computing provides services that are considered the basic building blocks of complex business applications. Services are platform-independent and can be discovered and composed dynamically. These features make services suitable to cope with the high rate of changes in business demands. 

The openness of large systems, the changes in environment conditions, and the partial knowledge of agents about the system require agents to have mechanisms that facilitate service discovery, self-organization of structural relations as service demand changes, and the promotion and maintenance of a cooperative behavior among agents to ensure the proper performance of the service discovery activity.

The main contribution of this PhD work is the proposal of a model for Open Service-Oriented Multi-Agent Systems. This model integrates agents that are located in a network as in a plain society, and agents that also are part of dynamic complex groups (i.e., virtual organizations). The model provides mechanisms for the management of virtual organizations and services provided inside them. In addition, the model provides a network structure based on homophily between agents that model an efficient decentralized service discovery to agents that only have a partial view of the system. Agents in this model have mechanisms that allow them to reason about their location in the network and the possible structural actions to improve their situation and the performance of the service discovery. Moreover, since decentralized service discovery relies on the cooperation of the agents, the framework provides mechanisms to facilitate the promotion of cooperation among agents through social plasticity and incentives. The proposed model and the mechanisms have been tested considering different configurations and compared with other approaches present in distributed systems.