Qualitative Reasoning - Advanced Artificial Intelligence - page 157

source temperature in ⑥ . In ⑦ , water boils which makes the pressure rise and
the boiling point improve. The process returns to ③ because the water
temperature is lower than the boiling point. In ⑧ , the process ends because the
water temperature reaches the heat source temperature. In ⑨ , the pressure in the
container is too high, which is described by process p-burst(CAN), and explosion
occurs.
4.5 Qualitative Simulation Reasoning
Kuipers published thesis titled “Common sense reasoning about causality:
deriving behavior from structure” in 1984. In this paper, a framework of
qualitative simulation reasoning is built, the qualitative structure and the
qualitative behavior express approach, which are abstracted from ordinary
differential equation, are briefly proposed. Subsequently, his thesis titled
“Qualitative Simulation” was published in AI magazine in 1986. In this paper,
the abstract relation is definite and a QSIM algorithm used for qualitative
simulation is proposed. The validity and incompleteness of the algorithm are
proved by abstract relation. These two papers have established the foundation of
qualitative simulation.
Qualitative simulation derives the behavior description from the qualitative
description of the structure. It directly uses the parameters of the part as state
variables to describe the physical structure. Qualitative constraints are obtained
directly from the physical law. The variation of a parameter with time is regarded
as the qualitative state sequence. The solving algorithm proceeds from the initial
state, generates each possible subsequent state, and then repeats this process with
consistent filtration until no new state appears.
The structure description of qualitative simulation is composed of systematic
state parameters and constraint relations. Parameters are regarded as the
differentiable function of time. Constraints are the binary or multi- relations. For
example, the derivative of speed is acceleration, which is expressed by
DERIV(Vel,acc). f = ma is expressed by MULT(m,a,f). The monotone increasing
of f with g is expressed by M + (f,g). The monotone decreasing of f with g is
expressed by M - (f,g).
Behavior description cares about the variation of parameters. Assuming
parameter f(t) is a differentiable function from [a, b] to [ ¯ , ¯ ], the landmark