Abstract This paper describes a computer program called PLR that derives the qualitative behavior of ordinary differential equations. Current qualitative reasoning programs derive the abstract behaviour of a system by simulating hand-crafted ‘qualitative’ versions of the differential equations that characterize the system and summarizing the results. PLR infers more detailed information by constructing and analysing piecewise linear approximations of the original equations. The analysis employs the phase space representation of dynamic systems theory. PLR constructs a phase diagram for a system of piecewise linear equations by partitioning phase space into regions in which the system is linear, analysing the linear systems, and combining the results. It pastes together the local analyses into a global phase diagram by determining which sequences of regions the trajectories can traverse. The current implementation of PLR only handles second-order systems, but the method extends to higher-order systems. As an example of PLR's performance, I present its analyses of the Lienard and van der Pol equations.
[1]
Johan de Kleer,et al.
Readings in qualitative reasoning about physical systems
,
1990
.
[2]
F. Brauer,et al.
The Qualitative Theory of Ordinary Differential Equations: An Introduction
,
1989
.
[3]
Michael L. Mavrovouniotis,et al.
Reasoning with Orders of Magnitude and Approximate Relations
,
1987,
AAAI.
[4]
M. Hirsch,et al.
Differential Equations, Dynamical Systems, and Linear Algebra
,
1974
.
[5]
Olivier Raiman.
Order of magnitude reasoning
,
1986,
AAAI 1986.
[6]
Elisha Sacks,et al.
Hierarchical Reasoning about Inequalities
,
1987,
AAAI.
[7]
Benjamin Kuipers,et al.
The Limits of Qualitative Simulation
,
1985,
IJCAI.