CSC 411: Artificial Intelligence
(Fall 2006)
Assignment 4: Chapters 8, 9
Due by November 30,
Thursday, 2006
- Expert systems: In Section 8.2 we
introduced the rule-based expert systems, especially production
systems. A toy example of production
systems is developed for diagnosing automobile problems, where 4
production rules are identified. To build a real-life expert system, the
development team may have multiple persons to extract and convert more
production rules and decide the system control strategy. For this end,
- Identify
possible knowledge engineers, domain experts, and potential end users for
such an application, and discuss the expectations, abilities, and needs
of each of these groups;
- Create
10 to 15 if … then … rules in English or pseudocode
(other than those prescribed in Section 8.2) to describe relations within
this domain, and create a graph to represent the relationships among
these rules (a and/or graph, refer Section 3.3);
- Recommend
and justify your answers to the following questions of choosing control
strategies: data-driven or goal-driven search? Breadth-first or
depth-first search? In what ways could heuristics assist the search?
- Planning: In Section 8.4 we
introduced planning and used blocks world example to develop the issues of
planning, including state representation and transition, atomic actions,
and frame problem. Carefully read this section and answer the following
questions:
- In
page 317 we created two frame rules 8 and 9. Create the remaining frame
axioms (rules) necessary for the four atomic actions pickup, putdown,
stack, and unstuck described in rules 4 through 7 in the same page;
- Use
the atomic actions and frame rules that you created to generate the
search space of Figure 8.19;
- Show
two more incompatible (precondition) subgoals
in the blocks world atomic actions of Figure 8.19.
- CF Algebra: In Section 9.2.1, we
introduced the Stanford Certainty Factor Algebra for uncertainty
information processing. Given the following rules in a “back-chaining”
expert system application:
A Ù not(B)
è
C (0.9)
C Ú D è E (0.75)
F è A (0.6)
G è D (0.8)
The system can conclude the
following facts (with confidences):
F(0.9)
B(-0.8)
G(0.7)
Use the Stanford certainty factor
algebra to determine E and its confidence.
- Fuzzy Sets: In Section 9.2.2 we
introduced the fuzzy sets and applications in control system design –
fuzzy controller. Carefully read this section and answer the following
questions:
- Continue
the inverted pendulum example in page 354 with two more iterations of the
controller where the output of one iteration is
the input value for the next iteration.
- Identify
another application area where fuzzy control might be appropriate.
Present a set of fuzzy rules for this domain.