|
|
|||||
|
In this project we establish a link between the classical system theoretical
framework of a continuous time state space and the world of discrete event
systems. This link is given by the algebraic concept of a field. In the continuous
world, typically, this field is the field of real or complex numbers. Restricting
our perspective on discrete event systems with a finite number of system variables
(e.g. finite state automata), so-called finite fields play an analogous,
important rôle. Apart from the plenty of analogies which can be drawn from
classical continuous system theory, additional propositions can be derived due to
the finiteness of the field. As a result, algebraic systems over finite fields
contain additional structure, of which, advantage can be taken most easily in the
linear case.
In linear circuit theory, the use of finite fields is common practice since the
early sixties. But nevertheless, most contributions left the input channels unused.
In our contributions, we interpret systems from circuit theory as subclasses of
finite state automata and link the inputs with the states by an adequate
feedback structure in order to impose specifications on the controlled automaton.
To this end, we adapt methods from continuous control theory on the finite field
scenario.
Current research on systems over finite fields concentrates on the following
aspects:
Modeling of Finite State Automata
Starting from the scratch or based on a given automaton model a state space model
over a finite field can be derived. In such a model, an automaton state is
represented by a state vector in a finite state space, similarly, events that
impose conditions on the state transition are represented by input vectors in a
finite input space. It can be shown that all transition relations, governing
admissible state transitions with respect to a given input vector, can be written
in a so-called Reed-Muller-Form. The Reed-Muller form is a canonical polynomial
representation, which exists for arbitrary functions over a finite field. Among
others, one issue of research is the development of efficient algorithms for
obtaining the Reed-Muller form of the transition relations for non-deterministic
automata.
System Analysis
Once a finite field model is available, it can be used for analyzing the
properties of the modeled discrete event system. Here, one major analysis task is
to determine the state space decomposition into invariant subspaces (induced by
the transition relation), which typically consists of cyclic states and/or of
tree states. In the case of linear systems, necessary and sufficient conditions
have been deduced that allow the determination of the entire state space
structure. In the nonlinear case, Gröbner-basis or principal ideal representations
for the respective nonlinear transition relations are appropriate means for an
efficient analysis - still work in progress.
Synthesis using Feedback
Provided that a system model over a finite field is controllable in some sense,
closing the loop by using state feedback is a (continuous world) standard method
for synthesizing desired system properties of the overall closed-loop system. If
the system model is linear then an image domain representation of the transition
relation, an adaption of the z-transform and of the polynomial approach, turns out
to be tailored for fixing closed-loop properties by specifying closed-loop system
invariants. Concerning nonlinear system models, a variety of research activities
is given by employing polynomial state feedback for setting the closed-loop
behavior.
This project was conducted by Hans Reger with support by the German National Research Foundation (Deutsche Forschungsgemeinschaft, DFG).
Reger, J., Schmidt, K.: A finite field framework for modelling, analysis and control of finite state automata, Mathematical and Computer Modelling of Dynamical Systems (MCMDS), vol. 10, issue 3-4, pp 253-285, Taylor & Francis, 2004. [PDF]
Reger, J., Schmidt, K.: Aspects on analysis and synthesis of linear discrete systems over the finite field GF(q), Proc. Euproean Control Conference ECC2003, Cambridge, United Kingdom, 2003. [PDF]
Reger, J.: Analysis of multilinear systems using gröbner-bases over the finite field GF(2), Proc. 4th International IMCAS Symposium on Mathematical Modelling (MATHMOD), Vienna, Austria, 2003.
Reger, J.: Cycle analysis for deterministic finite state automata, Proc. 15th IFAC World Congress, 2002. [PDF]