'r
V ’
V
=
a
o,
r (t) = {x(t),y(t),z(t)} (4)
This model can be applied in different operating
modes, employing as state vectors whether pseudoran
ge or carrier phase osservables, as collected by only
one receiver or in differential mode. Note that (4)
represents a costant acceleration model, in which
vectors v and a incorporates the linear and angular
components.
In order to take into account the error introduced by
our unique positioning device (GPS) we use following
formula:
s(t)=f{s(t)-t}+w(t) (5)
where
s(t) is the state vector;
f(t) describes mathematically the nonlinear
relationship between parameters in s(t)
and the process;
w(t) models sensor (rover receiver) syste
matic errors, which are considered as
white Gaussian noise with covariance Q.
In order to determine the components of s(t), kinematic
measurements are employed according to the formula
below:
z* =H-s(i) + n* (6)
where
z k measurements vector collected at discrete
time t k ;
H matrix of measuring states;
n k measurement Gaussian error with
covariance R.
cp(t) angle of tangent to trajectory at time t;
co(t) angular velocity of the van;
v(t) linear velocity;
a(t) linear acceleration;
wl(t), w2(t) Gaussian 2D positional error
components.
Since we are interested in planimetrie survey, we
consider only the (X,Y) GPS coordinates as computed
by differential correction in post-processing, therefore
the matrix form of equation (6) becomes:
k* =X k +n hk
[^2,* — yk y n 2,k
(8)
where
Z] k ) z 2 , k components ov measurements vector
at step k;
n i,k> n 2 , k Gaussian components of measurement
stochastic errors vector.
In this way the matrix of osservable states assumes
following form:
0 0
1 0
0 0 0
0 0 0
0\
being the state vector
s(i)={x(i) y{t) (pit) 00(f) v(t) a{t) ait)}
(9)
00)
4. STATE VECTOR ESTIMATE.
Since the Extended Kalman Filter (EKF) represents the
optimal estimate procedure in case of uncorrelated
measurements and Gaussian noise with null average,
the employed algorithm can be summarized by follo
wing formulas:
In more detail our nonlinear kinematic model is
composed by following 7 differential equations:
x(t) = vit)-COSi(p)
yit) = v(0 • sen(<p)
(pit) = coit)
< (bit) = ait) (7)
v(t) = ait)
ait) = 0 + w, it)
ait) = 0 + w 2 it)
where
x(t), y(t) 2D vehicle coordinates on mapping
frame E m ;
• Prediction step,
s(k + \\k) = ft(s(k\ky) (ll)
P(* + ll/t) = <P(k\k) ■ P(k\k) ■ ®(k\k) +Q(k) (12)
• Update step,
'A(k + 1) = H • P(k + 1 lit) • H'+R (13)
Uk +1) = P(k +11 *) • H'-A(* + 1)“' (14)
s(fc + llfc + l) = s(fc + llfc) + L(fc + l)-....
■[z(k + \)-U-s(k + l\k)] U5)