• Keine Ergebnisse gefunden

4 DEVELOPMENT OF THE CONCEPT

Im Dokument Production Engineering and Management (Seite 146-151)

The aim of this article is to create a theoretical concept of the communication between agents in a multi-agent system in order to simulate swarm intelligence. It is required that an object comes into the transportation area where it organizes its further transportation autonomously. The organization of transportation includes the communication between the object and the vehicles to pick the fastest transportation. The control of the route is not part of this article and concept.

4.1 Characteristics of agents in this concept

The agent traits can be described as targeted, self-organized and rational [21]. There are a few possibilities to solve the problem of figuring agents.

The three-tier architecture with working tiers for fast reaction, planning and modeling [22]. An agent with belief-desire-intention (BDI) architecture is segmented in his beliefs, which involve the status of his environment and his own, his desires, which influence the agent manner, and intentions to implement goals [23].

In order to generate decentralization and therefore flexibility, the communication between the agents has to be direct. This form of communication is called message-passing [24]. Another way, but not a highly decentralized one, would be the communication over a blackboard, where agents put information on this board and every other agent can read them [24].

There are different ways to find out the agent with the fastest transportation.

Two types should be presented: auction and negotiation. For an auction, agents submit tenders between them and a specific way can be chosen [25].

In negotiations, on the other hand, an agent gives another quotation which he can accept or decline [22].

4.2 Workflow of the concept

The concept is realized with two types of agents: agents who represent vehicles and agents who represent objects. The object agent can just solve the task to get his designated position, if he works together with a vehicle agent.

Figure 1 presents the concept of this application. For a better overview there are just three vehicle agents. It is viewable that the swarm intelligence lies with the vehicle agents. They communicate with each other to transport the object in the best way and solve the tasks together.

Figure 1: Workflow.

In the following part the sequence diagram is explained with its individual steps:

1. Inquiry: At the beginning the reader sends an inquiry on his frequency field.

2. Data communication: Given that there is an object with a tag in this field, the tag answers and begins with data communication to the reader [26]. Just due to the reader’s frequency field, the passive tag is able to send an answer [27].

3. Position: After that, the reader writes the cutrrent position on the tag.

4. New: During the data communication the agent will be transferred from the tag to the reader (agent-on-tag) [28]. The reader sends the agent to the agent-platform where it can run on [29]. The platform is responsible for the management and technical design of the agents [30].

5. Position comparison: The object agent will check if the current position is his designated position.

6. Inquiry: The positions aren’t equal. So the object agent has to ask the first vehicle agent A1 for transportation.

7. Index: In order to compare the different transportation times the vehicle agent A1 has to create an index.

8. The task of the index is to confirm the transportation effort for each vehicle. To succeed, the criteria driving time and energy status have to be included in the index.

9. The driving time means, on the one hand, the time the vehicle needs to reach the object. On the other hand, there might be the case that the vehicle is already transporting an object. In this case the driving time includes the time to finish this task and the time to reach the designated position. The energy status is important for the index, if it is so low that transportation is not possible. index to the agent with the next higher number A2.

12. Index: Agent A2 starts to create its own index and compares it with the index of the quotation.

13. Decline: The index of Agent A2 is better, so it declines the quotation.

14. Accept: Agent A1 accepts and gives the responsibility of the transportation to agent A2.

15. Quotation: Agent A2 starts a quotation with its index to the next higher agent.

16. Index: Agent A3 creates its own index.

17. Accept: Its index isn’t better so it accepts the quotation of agent A2.

18. If there were more agents, A2 would have to ask them.

19. Transportation request: The vehicle starts the transportation of the object.

20. Data communication: Before the tag is out of the field of the reader the agent is transferred from the platform to the reader.

21. Data and position: The reader transfers the agent and information back to the tag and changes the current position to a driving position.

22. Inquiry: At the end of the transport the reader at the final destination gathers the tag in its field.

23. Data communication: The tag answers the reader’s inquiry of the reader similar to point 2.

24. Position: The reader writes the current position on the tag.

25. New: The reader sends the information and the agent to the platform.

26. Position comparison: The agent compares the current position with its designation position. Because of the equality the object agent will not start an inquiry.

Another method to achieve the best transportation is the auction. In this case there are two possibilities; first a vehicle agent heads the auction and second, a collective agent works as representative of the swarm, while heading the auction.

4.3 Logical and physical challenges for concept implementation

There are different logical and physical challenges for implementation of this concept. The protocols of communication, the ontology and the speech have to fit with the requirements of the software architecture [31].

The software-agent-platform has to manage and control the agent and has to give agents information about other existing agents in the system.

In order to reach a very decentral concept, the automated guided vehicles should be equipped with small computers on which their agent can run. This requires a platform where agents on other hardware can communicate with each other.

The data inside the object tags should give information about the identification number, the loaded items, the agent and his data, the current and the designated position. The benefit of agent-on-tag is de-centrality but the high memory size required, brings the disadvantage that the time fields of the reader do not overlap.

5 BENEFITS FOR COMPANIES BY USING SWARM INTELLIGENCE

Im Dokument Production Engineering and Management (Seite 146-151)