Robotics Reference
In-Depth Information
decision of the expert to the various factors that caused him to reach that
decision.
The Inference Engine
The real strength of expert system programs is their ability to draw con-
clusions from premises. This ability is what makes an expert system intel-
ligent. The inference engine is the part of the expert system that knows
how to apply the rules in the knowledge-base and decide in which order
they should be applied when solving a particular problem. By interpret-
ing the rules in the knowledge-base, the system is able to draw its conclu-
sions. Two alternative strategies are available for making inferences from
the rules, called
forward chaining
and
backward chaining
.
A forward chaining inference engine reasons, from the premises, “for-
wards” to a conclusion. The process starts from the knowledge-base and
any data available as evidence. It first examines the current state of the
knowledge-base and evidence, then finds those rules whose premises can
be satisfied from known data (i.e., those rules in which the “
if
”partof
the rule is known to be true), and adds the conclusions of those rules
(i.e., the “
then
” parts) to the knowledge-base. With these conclusions
added, the system re-examines the complete knowledge-base and repeats
the process, which can now progress further because of its access to the
new information, until eventually, hopefully, the goal is reached.
Backward chaining works from the end (i.e., from the solution or goal
of the original problem), “backwards” in the sense that it tries to prove
the goal or conclusion by confirming the truth of all of its premises.
How Expert Systems Explain Their Reasoning
A chaining process will, when it has reached its conclusion, consist of
a chain of steps that can be traced by the expert system, enabling the
system to explain its entire reasoning processes to the user. In this way
the user can not only understand why the system reasons as it does, which
occasionally leads to the human expert learning a new idea or technique
from the system, (s)he can also detect flaws in the system, flaws that
might be traceable to a badly expressed rule or which might indicate that
a new rule needs to be added to the knowledge-base in order to plug a
specific gap in the system's knowledge.
The ability to explain their reasoning processes is a key feature of ex-
pert systems. They can justify their decisions and explain why they are
