Environmental Engineering Reference
In-Depth Information
5.1.6 Schedulers
Planning systems and scheduling systems work in the same domain, solve very
similar problems, and use the same core technologies. The primary difference is
that scheduling systems are more focused on detailed accounting of resources,
and they tend to generate complete and detailed plans.
5.2 Collaborative Languages
Most forms of collaboration require the collaborating agents to communicate
using a shared language. This language specifies the kinds of information that
can be exchanged between the agents, and in a real way, determines what one
agent can and cannot communicate to another. Well designed languages allow
the agents to say what is necessary to solve the problem. Poor languages limit
communication and support less than optimal solutions.
Computer systems support many kinds of communication. The simplest
communication, from a computer science point of view, is an exchange of mes-
sages whose content is an internal program data structure. The data structure
can be a simple primitive type like a number or string, or the structure can
be complex connections of records. This approach to communication is easy
to implement, and the communicating parties can immediately interpret the
content and meaning of the messages they exchange. However, this method
of communication has a limited expressive capability.
The other approach is to create a general purpose formal language. These
systems construct a computer language that is general in nature and able
to express a wide range of concepts. The first advantage of this approach is
flexibility, since these languages can express and collaborate on richer sets of
problems. Also, being a formal language, it can be documented and used by
different and disjoint teams building collaborative agents. Even though they
share no code or common ancestry, they can collaborate using this common
language. The Defense Advanced Research Projects Agency (DARPA) knowl-
edge sharing effort (KSE) has developed a very capable formal language for
intelligent systems to exchange information. In practice, both types of com-
munication are used in collaborative agents.
5.3 Reasoning with Partial Information
Intelligent behavior can be loosely broken down into problem solving (plan-
ning) and reasoning on evidence. Several technologies are used to reason on
partial information. This section will describe two common techniques: fuzzy
logic and Bayesian reasoning.
