Article
| Authors | Lars Braubach, Alexander Pokahr |
| Title | Method Calls Not Considered Harmful for Agent Interactions |
| Published in | International Transactions on Systems Science and Applications (ITSSA) |
| Medium | Volume 1/2 Number 7 |
| Date | November 2011 |
| Pages | 51-69 |
| Abstract | Since the very beginnings of agent technology it is considered a fundamental property that communication between agents is done using asynchronous message passing. One important reason for this assumption is that a method call on an agent would break its autonomy. Conceptually, it would allow an agent to exert control over the behavior of another agent by directly telling it to execute a specific method. Furthermore, more than one agent could call methods on the same agent concurrently possibly leading to state consistency problems. In contrast to agents, in mainstream programming paradigms like object, component or service orientation method calls are the dominant communication means. Being able to define interfaces from method signatures represents an important advantage of those approaches currently not available for agents. In this paper we will argue how it is possible to introduce service interfaces and method calls for agents without breaking their autonomy and endangering their state consistency. To achieve this, concepts from active objects, services and components are brought together. The usefulness of the approach is underlined with an example application from the area of distributed image rendering. |
| Document |  |
| Other formats | Din 1501 |
