In order to characterize the memory of a dynamical system, we use a concept known as state.
A system's state is defined as the minimal set of variables evaluated at
t=t0tt0
needed to determine the future evolution of the system for
t>t0tt0,
given the excitation
utut
for
t>t0tt0
.
We are given the following differential equation describing a system. Note that
ut=0
ut
0
.
ddtyt+yt=0
t1
y
t
y
t
0
(1)
Using the Laplace transform techniques described in the module on Linear Systems with Constant Coefficients, we can find a solution for
ytyt:
yt=yt0ⅇt0-t
y
t
y
t0
t0
t
(2)
As we need the information contained in
yt0yt0
for this solution,
ytyt
defines the state.
The differential equation describing an unforced system is:
d2dt2yt+3ddtyt+2yt=0
t2
y
t
3
t1
y
t
2
y
t
0
(3)
Finding the
qsqs
function, we have
qs=s2+3s+2
q
s
s
2
3
s
2
(4)
The roots of this function are
λ1=-1
λ1
-1
and
λ2=-2
λ2
-2
.
These values are used in the solution to the differential equation as the exponents of the exponential functions:
yt=c1ⅇ-t+c2ⅇ-2t
y
t
c1
t
c2
-2
t
(5)
where
c1c1
and
c2c2
are constants. To determine the values of these constants we would need two equations (with two equations and two unknowns, we can find the unknowns). If we knew
y0y0
and
ddty0
t1
y
0
we could find two equations, and we could then solve for
ytyt.
Therefore the system's state,
xtxt,
is
xt=ytddtyt
x
t
y
t
t1
y
t
(6)
In fact, the state can also be defined as any two non-trivial (i.e. independent) linear combinations of
ytyt
and
ddtyt
t1
y
t
.
Basically, a system's state summarizes its entire past. It describes the memory-side of dynamical systems.