Evaluation of Environment Contextual Services in Multiagent Systems - Agents and Artificial Intelligence - page 200

and addressed messaging is not efficient because of its genericity. Therefore, we also
re-implemented the broadcast and point-to-point communications using the same data
structures as the environment, in order to have a fair comparison between the environ-
ment and the broadcast solutions.
In the following, we compare the two last solutions. Firstly, we study the impact of a
variation in the number of agents. The results shown in Fig. 3 feature 10 to 100 agents.
The curve shows a clear advantage of the environment over the broadcast, and that the
more agents there are, the more EASI is comparatively interesting (from 27.8% for 10
agents to 53.9% for 100 agents).
Fig. 3. Comparative results of Broadcast and EASI in function of the number of agents
In the second phase (Fig. 4), we study the impact of a variation in the number of
updates, for 50 agents. During each update, all the agents move and in the case of the
Environment implementation, they modify their description accordingly. The run-time
gain of EASI over broadcast gets from 50.7% for 20 updates to 48.8% for 1000 updates.
Therefore, the cost of the update mechanism is not significant in comparison with the
difference in efficiency.
Another solution to reduce the cost of the matching process is to reduce the number
of entities that are tested for each message. For example, if the reception of messages is
conditioned by the skill of the agents, then we can cluster the agents in several commu-
nication environments according to this property value. This clustering should decrease
the cost of the matching for two reasons: the number of agents in each environment de-
creases, and the filter can be simplified by removing the test on the skill property. To test
this hypothesis, we have run experiments using several communication environments.
In these experimentation the following hypothesis have been done:
- We call Environment 3 interaction through three environments, and the agents are
grouped in function of their skill. The skill driver has been added in order to have
three skills.
- The agents send “request” messages. The agents interested in these messages are
those which are close to the sender and have the same skill. In experimentation
Broadcast , the messages are broadcasted to all the agents, while in Environment 3
they are managed by f recept 2 . This filter is designed after the filter f recept , without