Professional Documents
Culture Documents
Model Approach
He-Sheng Wang
Department of Communications and Guidance Engineering
National Taiwan Ocean University, Keelung 202, TAIWAN
Phone: +886-912266254, FAX: +886-2-24633492
E-mail: hswang@mail.ntou.edu.tw
ABSTRACT
The main purpose of this paper is to present a Markov process approach to the analysis of GNSS (Global Navigation Satellite
System) availability in Taiwan. The proposed model is capable of calculating a variety of statistical measures of the required
services at locations specified by the user. The primary input data for the availability model is the MTBF (Mean Time Between
Failure) and failure rate of the GPS (Global Positioning System) satellites. Then each visible set of satellites is checked to see
if it satisfies the service accuracy criterion. Finally, in order to meet the integrity requirement, the number of the satellites and
their corresponding geometry are verified to see if they meet the integrity threshold. For other types of navigation systems
configurations (such as GBAS, GRAS, or SBAS) different service criteria may be added in the model. The analysis addresses
GPS constellation of 24 satellites, and augmentation with geostationay satellites or pseudolites. Availability estimates for GPS,
SBAS and GBAS are obtained by software simulation.
1. INTRODUCTION
The advent of the Global Positioning System (GPS) has the potential to revolutionize aircraft navigation technique. Although
GPS alone has the ability to provide the primary means of navigation services for en-route, terminal and non-precision
approach phases of flight, augmentation is still needed to meet more stringent performance requirements on availability,
accuracy and integrity especially for the landing guidance during the precision approach phase. While the Satellite Based
Augmentation System (SBAS) is being implemented to provide seamless navigation guidance for all phases of flight down to
Category I decision height of 200 feet, the purpose of the Ground Based Augmentation System (GBAS) is aimed at furnishing
the aircraft with high accurate landing guidance during the precision approach phase and all the way down to touchdown.
In the perspective of US Federal Aviation Administration (FAA), the objective of using GPS is to provide enhanced
service and to reduce infrastructure cost for aircraft navigation. To be used as a primary-mean navigation system, the Required
Navigation Performance (RNP) for integrity, availability, accuracy and continuity must be met. Primary means of navigation
refers to the capability of planning an operation around scheduled outages so that the system is available for a particular flight
and the operational continuity, availability and accuracy requirements are met. The four parameters for the RNP are defined in
the following.
1. Accuracy: The degree of conformance between the estimated or measured position and/or velocity of a platform at
a given time and its true position and/or velocity.
2. Integrity: The ability of a navigation system to provide timely warnings to user when the system should not be
used for navigation. Values stated are the probability that a system does not have integrity.
3. Continuity: The probability that the signal supports navigation accuracy and integrity requirements for the duration
of the intended operation (approach or landing), given that the capability was available at the beginning of the
operation, The continuity requirement is expressed as a loss of continuity per unit of time.
4. Availability: The probability that the navigation and fault detection functions are operational and that the signal
accuracy, integrity, and continuity of function requirements are met. This is typically express as a fraction of time.
The main purpose of this paper is to present a Markov process approach to the analysis of GNSS (Global Navigation
Satellite System) availability in Taiwan. The proposed model is capable of calculating a variety of statistical measures of the
required services at locations specified by the user. The primary input data for the availability model is the MTBF (Mean Time
Between Failure) and failure rate of the GPS (Global Positioning System) satellites. Then each visible set of satellites is
checked to see if it satisfies the service accuracy criterion. Finally, in order to meet the integrity requirement, the number of the
satellites and their corresponding geometry are verified to see if they meet the integrity threshold. For other types of navigation
systems configurations (such as GBAS, GRAS, or SBAS) different service criteria may be added in the model. The analysis
addresses GPS constellation of 24 satellites, and augmentation with geostationay satellites or pseudolites. Availability
estimates for GPS, SBAS and GBAS are obtained by software simulation.
2. MATHEMATICAL PRELIMINARIES
In this section, we briefly review some materials needed in the following development. Subjects to be exploited include the
reliability theory and the descriptor systems theory.
The term Availability has its origin from the reliability theory. Therefore, to gain a further insight about how the
availability of navigation systems can be defined, it would be beneficial to review some of the reliability theory.
When an items function is time dependent and it is needed throughout a time interval, its reliability is known as mission
reliability, expressed as
R R (t )
where R(t) is the reliability defined as the probability that the component will operate correct until time t, given that it was
operational at time 0; t is the specified mission time duration. More useful reliability parameters for continuously operating
items are the mean time between failures (MTBF) and the availability. The MTBF tells us how frequently, on the average, we
can expect our item to experience an outage. The availability tells us the proportion of the time that we can expect our item to
be operating satisfactorily. Associated with these characteristics are the mean downtime (MDT), the average time that is takes
to return to an operating state after an outage has occurred, and the outage rate, the complement of availability, or the portion
of time that we can expect our item to be down.
For reliability evaluation, we will be using several functions for describing the failure process. The key functions are
1.
2.
3.
Q (t )
Q(t ) exp
( x )dx R(t )
0
(1)
The failure density of the component, f(t), can also be expressed in terms of reliability as
f (t )
dR(t )
dt
(2)
(t )
1
f ( )d
f (t )
R (t )
(3)
0, t 0
(4)
The choice of this density leads to a constant failure rate function (t ) . Of particular importance is the mean time to
failure (MTTF), given by:
MTTF
tf (t )dt
(5)
R(t )dt
(6)
In view of our previous definition, the simplest model of the system availability can be derived as follows:
A
MTBF
.
MTBF+MDT
However, for a complex system such as GNSS navigation system, this model appears to be too rough.
A complex system is composed of a given quantity of elements, any of which may be up or down or in a standby mode at
any point in time. The following general statements about the complex system can be made:
1.
At some point in time, say t=0, all elements are as designated to be. That is, elements designed to be normally
operating are operating, those designed to be normally in a standby, those designed to be performing a monitoring
function are doing so, etc.
2.
At any point in time, every element that is up (operating operable) will eventually be down.
3.
At any point in time, every element that is down will some time be up (restored or replaced).
4.
One-shot or cyclically operating elements can be said to be at t=0 in terms of a probability statement.
5.
Every combination of elements up, down, standby comprises a different system state.
6.
Upon a single failure or restoration event, the system passes form one state to another.
These statements describe a Markov process, the second type of Markov model. The states are discrete, defined by what
elements are up, down, standby. The time is continuous. Therefore the Markov model approach can be used for solving the
availability for complex systems.
Consider a general case consisting of a multitude of Markov success states, with transition among the states due to
failure or restoration events. Let ri j represent the rate from state j to state i. The transition rate diagram is shown in Figure 1.
Let us define all possible success states 1, 2,, N and fail state F. Let Pi (t ) be the probability of being in state i (i=1, 2,, N,
F) at time t. Hence Pi (t t ) is the probability being in state i at time (t+t). Suppose that all states are mutually exclusive,
then we have
N
P (t ) P (t ) 1.0
i
(7)
i 1
Define a time increment t that is so small that the probability of more than one event taking place during t is negligible. The
probability of being in a particular state i at time (t+t) is the probability that
1.
2.
The system is in some other state at the time t, and it goes from that state into state i during t.
Events 1 and 2 are mutually exclusive, so Pi (t t ) is the sum of the probability of event 1 plus the probability of event
2. Event 1 is the combination of two independent
ki
all k i
system is in some state other than i at start of the interval times the probability that it goes from that state into state i during the
interval t. Therefore the probability of event 2 is Pj (t ) ( ri j t ) . Hence, the probability of being in state i at the time (t+t) is
Pi (t t )
P (t ) [r
j
ij
t ] Pi (t ) [1
all j i
Pi (t t ) Pi (t )
P (t ) [r
j
all j i
ki
t]
(8)
all k i
ij
t ] Pi (t )
all k i
ki
(9)
Pj (t ) ri j Pi (t )
rki (10)
t
all j i
all k i
Allow the notation rii to represent the sum of all transition rates out of state i:
r11 r21 r31 L rN 1
r22 r12 r32 L rN 2
M
rii
(11)
ki
all k i
rNN
M
r1N r2 N L rN -1, N
rij Pj (t ) rii Pi (t )
t
all j i
(12)
Upon letting t approaches zero, Equation (12) then takes the following differential form:
lim
t 0
Pi (t t ) Pi (t )
t
d
Pi (t )
rij Pj (t ) rii Pi (t )
dt
all j i
(13)
(14)
Hence, the availability at time t is the sum of the probabilities of being in each of the success states 1, 2, 3,, N at time t.
In order to find the N values of Pi (t ) from the N differential equation, we must have a boundary condition. Without loss
of generality, we can assume that
P1 (0) 1.0
Pi (0) 0, i 1
We can also solve A(t ) with the system starting in any state i, not necessary state 1. A more general boundary condition
is
Pj (0) 1.0
Pi (0) 0, i j
So the exact solution for the availability of a complex system with N success states over the time t is
N
A(t )
P (t )
i
(15)
i 1
where the Pi (t ) values are the solutions to the set of N differential equations
d
P1 (t ) r11 P1 (t ) r12 P2 (t ) r13 P3 (t ) L r1N PN (t )
dt
d
P2 (t ) r21 P1 (t ) r22 P2 (t ) r23 P3 (t ) L r2 N PN (t )
dt
d
P3 (t ) r31 P1 (t ) r32 P2 (t ) r33 P3 (t ) L r3 N PN (t )
dt
(16)
M
d
PN (t ) rN 1 P1 (t ) rN 2 P2 (t ) rN 3 P3 (t ) L rNN PN (t )
dt
P1 (0) 1.0 and Pi (0) 0 for all i 1
In matrix format the availability of a complex system over time t can be obtained through the solution of the following state
space equation:
d
dt P1(t )
r11
d P 2(t )
r 21
dt
r 31
rN 1
PN ( t )
dt
r12
r 22
r 32
rN 2
r13
r 23
r 33
rN 3
r1N
P1(t )
r 2 N
P 2(t )
r 3 N
PN (t )
rNN
Note that, because the Pis constitute the complete probability space, it is necessary that the following condition is satisfied for
all t
P (t ) P (t ) P (t ) L
i
PN (t )
(18)
i 1
Equations (17) and (18) constitute a descriptor system. Before introducing the elementary facts about descriptor systems,
we first consider a degenerate case of Equations (17)-(18) in the following. Equation (17) describes the instantaneous
availability. There is another type of the system availability, namely the steady-state availability. To obtain the steady-state
availability, we simply take the limit of equation (17) as t approaches infinite:
Pi lim Pi (t )
t
lim
d
Pi (t ) 0
dt
(19)
(20)
where the values of Pi are the solutions to the set of independent, linear equations:
(21)
To avoid a trivial solution, for the most trivial equation (usually the last one) in the set we can substitute
N
P P
P2 P
(22)
i 1
Equation (22) is valid because all states are mutually exclusive and exhaustive. Equations (21) and (22) can be put in a more
compact form as follows:
0
0
0
1
r12
r 22
r 32
r13
r 21
r 31
r( N 1), 2
r( N 1),3
r11
r 21
r 31
r
( N 1),1
r1N
r 2 N
r 3 N
L r( N 1), N
1
P1
P 2
(23)
N
P
The derivation of Equations (18) through (23) in the preceding paragraph demonstrates that steady-state availability is
A lim A(t)
t
Pi
all sucess
states
(24)
Let us now get back to the solution of the equations (17)-(18) which can be put in the following descriptor form:
& t ) FP(t ) Bu(t )
EP(
(25)
where
1
0
E M
0
0
r11
r 21
r 31
r( N 1),1
0
1
M
0
0
L 0
0 L
O M
L 1
L 0
0
P1 (t )
P (t )
, P(t ) 2 ,
M
M
PN (t )
0
r12
r 22
r 32
r13
r 21
r 31
r( N 1),2
r( N 1),3
r1N
r 2 N
r 3 N
, B
L r( N 1), N
M
,
0
and u(t) is the unit step function. Equation (25) is different from the ordinary differential equations in that the matrix E is
singular, in general. We have the following theorem concerning the solvability of the descriptor systems.
Theorem 1. The descriptor system (25) has a unique solution for a given input function u(t) and any initial condition of the
form EP(t0) if and only if the matrix pair {E, F} is regular, i.e., there exists a scalar 0 such that 0E-F is not identically zero.
The last equation in (25) describes an algebraic constraint among the state variables and the input function, which is the
key to the analysis of GNSS availability. Specifically, if we want to develop different Markov models for various GNSS
augmentation systems (such as SBAS, GBAS, GRAS, etc.), we simply add a set of appropriate algebraic constraints to the
descriptor systems, and construct a set of input functions so that solution trajectories of Equation (25) can be arbitrarily
assigned. This concept exactly coincides with the notion of the controllability of the descriptor systems.
Theorem 2. The descriptor system (25) is completely controllable if and only if rank[sE-F B]=N for all complex scalar
numbers s.
3. MARKOV MODEL APPROACH TO SOLVING COMPLEX SYSTEM AVAILABILITY
The general information flow and computation diagram is depicted in Figure 2. This kind of computational diagram was first
proposed by Poor [4]. Before we can use the Markov model to compute the system availability, we need to derive an effective
system MTBF in order to reflect the fact that, for different satellites, MTBF should not be the same, in general.
Recall our earlier results:
MTBF
(26)
A(t )dt
Let t be the time required for the system to get from state 1 at time t=0 to the absorbing state F. The mean of the variable t,
which is the MTBF,
d
MTBF=E t
M t ( )
dt
where M t ( ) E e t
(27)
0
e t f (t )dt is the moment- generating function. Notice that f(t) is the failure density function for
et f (t )dt
(28)
e st f (t )dt
(29)
d
d
M t ( ) | 0 L
d
ds
f (t )
s 0
(30)
dA(t )
dt
(31)
(32)
Hence,
f (t )
dP (t )
dA(t )
F
dt
dt
(33)
MTBF
d
L
ds
dPF (t )
dt
(34)
s 0
d
Pi (t ) sFi ( s ) Pi ( 0)
dt
(35)
P1 (0) 1.0
(36)
Pi (0) 0 , i 1
Hence,
d
P1 (t ) sF1 ( s ) 1
dt
d
L
Pi (t ) sFi ( s ) , i 2, 3,K , N , F
dt
L
(37)
d
L
ds
dPF (t )
dt
s 0
d
sFF ( s )
ds
s 0
(38)
To serve as a illustrated example, we consider here a basic scenario of the availability model. We consider a 24 GPS
satellite constellation with the following parameters:
The system reliability diagram is shown in Figure 3, and the corresponding success states and failure state is shown in Table 1.
The computed availability for a single time epoch is:
Availability (GPS) = 0.9984632074.
4. SIMULATION RESULTS
In the section, simulation results for SBAS and GBAS systems availability are given. The main parameters are the same as in
the previous section. Basic assumptions are similar to GPS Risk Assessment Study reported by Johns Hopkins University [1].
For SBASE case, a 24 satellites constellation plus a geostationary satellite that has full coverage of Taipei Information Region
(FIR) is assumed to-
,
,
,
5 of 8 require
,
Figure 3. Block Diagram for GPS Availability Analysis
gether with a possible deployment of the ionospheric grid points. At the beginning of the simulation, the in-view satellites and
their ECEF coordinates are given in Figure 4 and Table 2, respectively.
------------------------------------------------------------Success Fail
Items
Items Items under
State State
UP
Down
Restoration
--------------------------------------------------------------------------01
08
00
0
02
07
01
0
03
06
02
0
04
05
03
1
**************************************************
*
05
06
07
04
03
02
04
05
06
1
2
2
08 01
07
2
09 00
08
2
--------------------------------------------------------------------------Table
1. Success State and Failure State Definition
--
The geometric dilution of precision (GDOP) of the in-view satellites is incorporated into the descriptor system (25) as a
weighting matrix to partly reflect the accuracy requirement for GNSS navigation system. For a given set of in-view satellites,
the weighting matrix can be computed as follows. Assume the visible satellites are given as in Table 2. We first compute a
vector of partial GDOPs. The results are shown in Table 3. Then a weighting matrix can be derived as follows:
PRN 3
-1.8142e+007
-1.5163e+006
1.9313e+007
PRN 8
5.5144e+006
1.641e+007
1.985e+007
PRN 11
-1.3535e+007
2.2529e+007
3.1785e+006
PRN 13
5.6225e+006
2.4257e+007
-9.391e+006
PRN 16
-2.6473e+007
-1.8316e+006
-1.3719e+006
PRN 19
-1.1716e+007
1.0497e+007
2.1513e+007
PRN 27
-8.2794+004
2.2177e+007
1.4764e+007
PRN 31
1.4707e+007
2.1864e+007
-1.7271e+006
WGDOP =
diag[GDOP1, GDOP2, GDOP3, GDOP4, GDOP5, GDOP6, GDOP7, GDOP8] = diag[1.9462,
GDOP1
PRN 3111316192731
GDOP2
PRN 381316192731
GDOP3
PRN 381116192731
GDOP4
PRN 381113192731
GDOP5
PRN 381113162731
GDOP6
PRN 381113161931
GDOP7
PRN 381113161927
GDOP8
1.9462
1.8911
2.7357
1.8862
2.1463
1.8364
1.8732
2.0776
Table 3. Partial GDOPs: Each of GDOPi is Computed by Excluding One Satellite from the Observed Set
given in terms of the protection levels which are shown in Figure 8 in the next page.
Availability analysis for GBAS model is given in Figure 7. A world-wide GBAS availability simulation is also
performed. For each location, number of the satellites in-view and the coordinates of the satellites are computed by using the
software package Satellite Navigation TOOLBOX for Matlab, by GPSoft LLC. The result is shown in Figure 9 in the next
page and also summarized in Table 4.
5. CONCLUSIONS
In this paper, we have proposed a Markov process approach to the analysis of GNSS availability. Both SBAS and GBAS
constellations are considered. Based on the assumptions similar to [1], the results are also compatible with it.
No.
City
Latitude
Longitude
Availability
London
Liberia
52N
0W
0.99999
7N
10W
0.9999
Iceland
65N
22W
0.9999
Buenos Aires
30S
58W
0.99999
Ecuador
3S
80W
0.9999
Winnipeg
50N
95W
0.99999
Los Angeles
34N
118W
0.99999
North Alaska
70N
150W
0.9999
Honolulu
22N
158W
0.9999
10
Ross Sea
75S
180W
0.9999
11
Tokyo
36N
140E
0.9999
12
Perth
32S
115E
0.99999
13
Singapore
2N
75E
0.9999
14
Indian Ocean
45S
60E
0.9999
15
Aral Sea
45N
50E
0.99999
16
Cape Town
35S
18E
0.9999