Information Technology Reference
In-Depth Information
500 Advanced Artificial Intelligence
(1) Improve problem solving capacity. Because of the distributed characteristics
of the intelligent system, its problem solving capacity increases substantially.
First of all, once high reliability, communications path, processing nodes, as
well as knowledge of the redundancy are failure, the whole system just
reduces the response time or solving accuracy and will not completely
paralyzed; Second, the system is easy to expand. To increase processing units,
it can expand the system ' s scale and increase problem solving ability.
Thirdly, modules characteristics will enable the design of the whole system
very flexible.
(2) Improve the efficiency of problem solving. Because the nodes of the
distributed intelligent system can be parallel to solve problems, we can
develop parallelization to solve problems and increase the efficiency of
problem solving.
(3) Expand the scope of application. Distribution of intelligent technology can
break the restriction of current knowledge engineering field through only
using an expert. In distributed intelligent systems, different areas, and even
different experts from the same field can collaborate to solve a problem that
particular experts can not. At the same time, many non experts can cooperate
to solve the problem which may also meet or exceed an expert level.
(4) Reduce the complexity of the software. Distributed intelligent system will
broke down a task into sub tasks to solve a number of relatively independent
of the sub tasks. The result is to reduce the complexity of the problem solving.
Distributed intelligent research can be traced back to the late 1970s. Early
distributed intelligent research focuses on distributed problem solving (DPS).
Their goal is to create large size of the cooperative groups to work together to
solve a problem. In distributed problem solving systems, data, knowledge,
control systems are distributed in the various nodes. There is no overall control
and knowledge of the overall data and storage. Because no system in the node
has enough data and knowledge to solve the whole problem, each node needs to
exchange information, knowledge, and problem solving state. They are through
mutual cooperation for solving complex problems of collaboration. In a pure
DPS system, the problem is broken down into tasks. For solving these tasks, we
need to design a specific task execution system for the problem. All the
interactive strategies have been integrated for part of the overall system design.
This is a top-down design of the system, because the system is designed to meet
the needs of top requirements.
Hewitt and his colleagues developed an ACTOR model based concurrent
programming design system (Hewitt,1983). ACTOR model provides parallel
Search WWH ::




Custom Search