Environmental Engineering Reference
In-Depth Information
environment and how they should react to new stimuli. Three states are obvious
for the agents:
1. Active agents are agents that observe everything of their fields, and fully
accept the input and influence of other agents according to the six laws.
In other words, active agents behave as expected from the design section.
2. Sleeping agents are agents that no longer possess intentions to move, and
that other agents no longer can expect to influence. However, stimuli from
the environment and input from other agents can grow so strong that the
agent is awakened again.
3. Locked agents are agents that no longer are under influence of the environ-
ment or other agents, but internally can still decide to unlock and become
active again.
The interesting issues are the transitions from state to state. For a sleeping
agent, there are two options: The agent itself or its group expects that the agent
can improve its position, or the agent is being pushed by the environment. An
agent could expect to improve its position if free space above has become avail-
able, and not necessarily directly above the agent. This could also happen if a
larger chunk of free space has become available earlier in the time domain.
The pressure from other agents can be controlled by a threshold value, so the
agent is awakened if the forces applied from other agents are too high, given by
the summed force of impact from others:
F I O
i F i S i d i
where F i is the force from the i th agent that want the position, S i is the satisfaction
rate of the i th agent, and d i its distance in time to the sleeping agent.
Also, the pressure on an agent from its group can be expressed as a summing
force that can break a threshold value and bring the agent awake again:
=
j F j 1
F I G =
d j
where F j is the applied force from the j th agent in the group and d j its distance
in steps to the sleeping agent. Given those transitions, an agent could fall asleep
if it is has found a steady state and is not in conflict with other agents. It could
also fall asleep if its group has found a stable level, but some members are
oscillating.
13.2
Predecessor Validation
An agent adjusts its position according to the free space around it, but also under
influence of its predecessor's position. Experiments have shown that especially
during the initial settling time of an agent group, some agents were strongly
influenced by their predecessor agents, which were not very reliable with respect
to their final position.
 
Search WWH ::




Custom Search