armin@ai.toronto.edu (Armin Haken) (02/12/88)
There will be an AI seminar on Tuesday 23 February at 2PM in room SF 1105, given by Dr. Elisha Sacks of MIT. The abstract follows. Mr. Sacks is a candidate for a faculty position. Hosting is Hector Levesque. ----------------------------- Automatic Qualitative Analysis of Ordinary Differential Equations Using Piecewise Linear Approximations by Elisha Sacks This talk explores automating the qualitative analysis of physical systems. Scientists and engineers model many physical systems with ordinary differential equations. They deduce the behavior of the systems by analyzing the equations. Most realistic models are nonlinear, hence difficult or impossible to solve explicitly. Analysts must resort to approximations or to sophisticated mathematical techniques. I describe a program, called PLR (for Piecewise Linear Reasoner), that formalizes an analysis strategy employed by experts. PLR takes parameterized ordinary differential equations as input and produces a qualitative description of the solutions for all initial values. It approximates intractable nonlinear systems with piecewise linear ones, analyzes the approximations, and draws conclusions about the original systems. It chooses approximations that are accurate enough to reproduce the essential properties of their nonlinear prototypes, yet simple enough to be analyzed completely and efficiently. PLR uses the standard phase space representation. It builds a composite phase diagram for a piecewise linear system by pasting together the local phase diagrams of its linear regions. It employs a combination of geometric and algebraic reasoning to determine whether the trajectories in each linear region cross into adjoining regions and summarizes the results in a transition graph. Transition graphs explicitly express many qualitative properties of systems. PLR derives additional properties, such as boundedness or periodicity, by theoretical methods. PLR's analysis depends on abstract properties of systems rather than on specific numeric values. This makes its conclusions more robust and enables it to handle parameterized equations transparently. I demonstrate PLR on several common nonlinear systems and on published examples from mechanical engineering. -- || Armin Haken armin@ai.toronto.edu || || (416)978-6277 ...!utcsri!utai!armin || || UofT DCS, Toronto M5S 1A4 CDN armin%ai.toronto@csnet-relay ||