Autonomous role discovery for collaborating agents
Software—Practice & Experience
| Hi-index | 0.00 |
Roles represent a great model to deal with interactions and sociality of autonomous entities like agents, and in fact in order to ease their adoption several role approaches have been developed in the Role Based Collaboration (RBC) field. Some of the main difficulties in designing and developing agent role based approaches are the needs of providing a good dynamism and a good perception of the played role. The former (dynamism) is the capability of an agent to assume, use and release a role at run-time; while the latter (perception) is the capability of an agent to perceive the role played by another agent without having to explicitly query such agent or the role environment about. While dynamism has been achieved with several techniques, the role perception is more difficult to reach and often requires deep changes in agent structures, like class refactoring. Other difficulties arise when the agent is masqueraded, for security reasons, by a proxy. This paper presents a role approach that enables Java agents to dynamically play and perceive roles. This approach exploits a dynamic class refactoring performed on the fly, in order to make visible and perceivable assumed roles.