Comparison of the different solutions for coordination
Some comparison criteria:
- what is the basic coordination mechanism?
- is the coordination inside or outside the agents? Inside => to change the protocol, the code of all agents need to be changed
- what does the agent need to know before to participate? (language, architecture)?
- how to find other agents?
- is the protocol explicitly described somewhere for the agents to "learn" the protocol?
- is it centralized?
- what is the performance?
| oriented to | mechanism | coordination | requirements | acquaintances | explicit protocol | centralized | performance |
|---|---|---|---|---|---|---|---|
| agent | ACL - simple | inside | ACL | YP | no | no | 2p + p2 messages1 |
| environment | action/perception - simpler | mixed2 | Cartago | YP | no | no3 | 2+p actions + p2 perceptions |
| organization | schemes - complex | outside | Moise | by roles | yes | no3 | idem env |
| + org actions + org perception | |||||||
Remark: the organization is better for open systems, however it costs something (complexity of the application and performance).
ACL = Agent Communication Language YP = Yellow pages
p = number of participants