Environmental Engineering Reference
In-Depth Information
large part of the world model and represents the use and control of resources,
is external to the agents performing the scheduling.
In other domains, the individual agents have direct control of resources and
often they do not share much of their world model with other agents. Such
agents tend to have a high degree of individual identity and when cooperat-
ing, the individual agents must negotiate as peers to achieve their common
objectives. Their self-model is expansive and it represents the status of all
resources and systems under their control. Spacecraft are agents with a high
degree of individual identity since they manage unique resources in unusual
places. For example, multiple spacecraft attempting to perform joint science
would need to coordinate their positions, their orientations, and the times
when they need to perform actions such as data collection.
In some domains, some goals can be achieved with only sporadic coopera-
tion, while others require continuous contact as the execution proceeds. The
previous spacecraft scenario is a good example of sporadic cooperation to col-
lect the necessary science. High precision formation flying is an example of
continuous cooperation since each spacecraft must constantly sense and mod-
ify its position in relationship to its neighbors. Some domains have a specific
hierarchy of responsibility, with the lower agents subservient to the upper.
Human controllers of a spacecraft demonstrate this hierarchical cooperation.
The controllers make the high level decisions, which are communicated to the
spacecraft as the lower level agent for execution. In other domains, agents are
direct peers who work together to come to a common agreement.
Finally, agents differ in how well they learn from experience. Some systems
have fixed, prescribed rules to specify how the agent should operate under all
known situations. Most current spacecraft fit in this category. Other systems
learn as they operate. These systems can adapt to changing environments,
and over time, can become more skilled. Human agents are a good model of
learning agents. With the attributes of autonomous agents described, it is now
possible to examine the behavior of groups of autonomous agents. This is the
topic of the next section.
7.2.2 Agent Cooperation
There are a number of software-related aspects of cooperative autonomy that
are embodied in agent technologies, which is a broad field. Figure
7.2
(see
Chap.
5
) shows an overview of agent technologies and the lower level tech-
nologies that are used to construct agents.
Computer-based agent technology is an active area of research and its goal
is to build computerized autonomous agents that fit within the models defined
simple case, agents work alone to achieve their goals. These types of agents
usually assume that no other agent is in the environment, and their plans
can be generated and pursued without concern over interference from others.
Sometimes, multiple agents work cooperatively to achieve a common goal.
