SlideShare a Scribd company logo
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
DOI : 10.5121/ijccms.2013.2201 1
COMPUTATIONAL COMPLEXITY COMPARISON OF
MULTI-SENSOR SINGLE TARGET DATA FUSION
METHODS BY MATLAB
Sayed Amir Hoseini1
and Mohammad Reza Ashraf 2
1
Department of Electrical Engineering, Amirkabir University of technology, Tehran, Iran
a.hoseini@aut.ac.ir
2
Department of Electrical Engineering, University of Tehran, Tehran, Iran
m.r.ashraf@ut.ac.ir
ABSTRACT
Target tracking using observations from multiple sensors can achieve better estimation performance than
a single sensor. The most famous estimation tool in target tracking is Kalman filter. There are several
mathematical approaches to combine the observations of multiple sensors by use of Kalman filter. An
important issue in applying a proper approach is computational complexity. In this paper, four data fusion
algorithms based on Kalman filter are considered including three centralized and one decentralized
methods. Using MATLAB, computational loads of these methods are compared while number of sensors
increases. The results show that inverse covariance method has the best computational performance if the
number of sensors is above 20. For a smaller number of sensors, other methods, especially group sensors,
are more appropriate..
KEYWORDS
Data fusion, Target Tracking, Kalman Filter, Multi-sensor, MATLAB
1. INTRODUCTION
Data fusion is the process of combining information from a number of different sources to
provide a robust and complete description of an environment or process of interest. Data fusion
is of special significance in any application where large amounts of data must be combined,
fused and distilled to obtain information of appropriate quality and integrity on which decisions
can be made. Data fusion finds application in many military systems, in civilian surveillance and
monitoring tasks, in process control and in information systems. Data fusion methods are
particularly important in the drive toward autonomous systems in all these applications [1].
Estimation is the single most important problem in sensor data fusion. Fundamentally, an
estimator is a decision rule which takes as an argument a sequence of observations and whose
action is to compute a value for the parameter or state of interest. Almost all data fusion
problems involve this estimation process: we obtain a number of observations from a group of
sensors and using this information we wish to find some estimate of the true state of the
environment we are observing. Estimation encompasses all important aspects of the data fusion
problem. The most famous estimation tool in target tracking is Kalman filter.
Tracking targets with Kalman filtering is an active research area and there are substantial
literatures in this field such as [2], [3] and [4]. [1,4] talk about several multi-sensor data fusion
methods. [5] and [6] are also the same but latest attempts. In [1], the computational complexities
of these methods are compared based on mathematical formulation, but no simulation or
implementation proof is presented. In this paper, the aim is to continue the related work in [1]
and evaluate its claim using MATLAB simulations.
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
2
This paper begins with a brief summary of the Kalman filter algorithm. The intention is to
introduce notation and key data fusion concepts; Prior familiarity with the basic Kalman Filter
algorithm is assumed. The multi-sensor Kalman filter is then discussed. Four main algorithms are
considered; the group-sensor method, the sequential sensor method, the inverse covariance form
and the track-to-track fusion. For the related concepts and additional formulation refer to [1].
After that, in section 3, these methods are evaluated in theory and simulation. Finally, Simulation
results and a comparison of these methods are presented in section 4.
2. KALMAN FILTER IN TARGET TRACKING
The Kalman Filter is a recursive linear estimator which successively calculates an estimate for a
continuous valued state, that evolves over time, on the basis of periodic observations that of this
state.
2.1. State and Observation Models
The starting point for the Kalman filter algorithm is to define a model for the states to be
estimated in the standard state-space form and observation model for data received from sensor
[7]. A general motion model used in discrete Kalman filter for target tracking is:
( ) ( ) ( 1) ( ) ( )x k F k x k G k v k= − + (1)
( ) ( ) ( ) ( )z k H k x k w k= + (2)
where F(k) is the state transition matrix, x(k) is the state vector at time k, G(k) is the process
noise gain matrix, The process noise v(k) and the measurement noise w(k) are zero mean,
mutually independent, white, Gaussian with covariance Q and R respectively. z(k) is the
measurement vector at time k and H(k) is observation matrix of the states computed at time k.
{ } { }( ) ( ) 0, ,E v k E w k k= = ∀ (3)
{ } { }( ) ( ) ( ), ( ) ( ) ( ).T T
ij ijE v i v j Q i E w i w j R i = = (4)
2.1.1. State Prediction
The state and state covariance matrix at time k-1 are predicted to time k as follows:
ˆ ˆ( | 1) ( ) ( 1| 1)x k k F k x k k− = − − (5)
( | 1) ( ) ( 1| 1) ( ) ( ) ( ) ( ),T T
P k k F k P k k F k G k Q k G k− = − − + (6)
where ˆ( 1| 1)x k k− − is the estimated state vector at time k, ( 1| 1)P k k− − is the estimated state
covariance matrix at the same time, ˆ( | 1)x k k − is the predicted state and ( | 1)P k k − is the
predicted state covariance matrix.
2.1.2. Measurement update
At time k an observation ( )z k is made and the updated estimate ˆ( | )x k k of the state ( )x k ,
together with the updated estimate covariance ( )|P k k is computed from the state prediction
and observation according to:
ˆ ˆ ˆ( | ) ( | 1) ( )( ( ) ( ) ( | 1))x k k x k k W k z k H k x k k= − + − − (7)
and
( | ) (1 ( ) ( )) ( | 1)(1 ( ) ( )) ( ) ( ) ( )T T
P k k W k H k P k k W k H k W k R k W k= − − − + (8)
where the gain matrix ( )W k is given by:
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
3
1
( ) ( | 1) ( ) ( ) ( | 1) ( ) ( )T
W k P k k H k H k P k k H k R k
−
 = − − +  (9)
2.2. The Multi-Sensor Kalman Filter
Many of the techniques developed for single sensor Kalman filters can be applied directly to
multi-sensor estimation and tracking problems. In principle, a group of sensors can be considered
as a single sensor with a large and possibly complex observation model. In this case the Kalman
filter algorithm is directly applicable to the multi-sensor estimation problem. This technique is
called “group-sensor method”. However, as will be seen, this approach is practically limited to
relatively small numbers of sensors.
A second approach is to consider each observation made by each sensor as a separate and
independent realization, made according to a specific observation model, which can be
incorporate into the estimate in a sequential manner. Again, single-sensor estimation techniques,
applied sequentially, can be applied to this formulation of the multi-sensor estimation problem.
This technique is called “sequential-sensor method”. However, as will be seen, this approach
requires that a new prediction and gain matrix be calculated for each observation from each
sensor at every time-step, and so is computationally very expensive.
A third approach is to explicitly derive equations for integrating multiple observations made at
the same time into a common state estimate. Starting from the formulation of the multi-sensor
Kalman filter algorithm, employing a single model for a group of sensors, a set of recursive
equations for integrating individual sensor observations can be derived. This method is called
“inverse covariance form”.
The systems considered to this point are all ‘centralized’; the observations made by sensors are
reported back to a central processing unit in a raw form where they are processed by a single
algorithm in much the same way as single sensor systems. It is also possible to formulate the
multi-sensor estimation problem in terms of a number of local sensor filters, each generating
state estimates, which are subsequently communicated in processed form back to a central fusion
center. This distributed processing structure has a number of advantages in terms of modularity
of the resulting architecture. However, the algorithms required to fuse estimate or track
information at the central site can be quite complex. This Fusion method is called “track-to-track
fusion” and discus as a fourth approach.
In the future equations, Indexes indicate the corresponding sensor number. For example, ( )iz k is
the observation of ith
sensor at the kth
time step.
2.2.1. The Group-Sensor Method
The simplest way of dealing with a multi-sensor estimation problem is to combine all
observations and observation models in to a single composite ‘group sensor’ and then to deal
with the estimation problem using an identical algorithm to that employed in single-sensor
systems. The combinatorial observation vector is defined as:
1( ) ( ),..., ( ) ,
TT T
sz k z k z k   (10)
and combinatorial observation model is as:
1( ) ( ),..., ( )
TT T
sH k H k H k   (11)
and
1( ) ( ),..., ( ) ,
TT T
sw k w k w k   (12)
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
4
and
{ } { }
{ }
1 1
1
( ) ( ) ( ) ( ),..., ( ) ( ),..., ( )
( ),..., ( ) ,
TT T T T T
s s
s
R k E w k w k w k w k w k w k
blockdiag R k R k
   = = Ε =   
=
(13)
Observation noise covariance is as a block-diagonal matrix which each block of diagonal is the
observation noise matrix of each sensor. Observation vector are made in a combinatorial manner.
Predication equations are like single sensor Kalman filter (equations (1) and (2)).
2.2.2 The Sequential-Sensor Method
A second approach to the multi-sensor estimation problem is to consider each sensor observation
as an independent, sequential update to the state estimate and for each observation to compute an
appropriate prediction and gain matrix. The final formulation provide below:
ˆ ˆ( | 1) ( ) ( 1| 1)x k k F k x k k− = − − (14)
( | 1) ( ) ( 1| 1) ( ) ( ) ( ) ( ).T T
P k k F k P k k F k G k Q k G k− = − − + (15)
( ) ( ) ( | , 1) ( ) ( )T
p p p pS k H k P k k p H k R k= − + (16)
1
( ) ( | , 1) ( ) ( )T
p p pW k P k k p H k S k−
= − (17)
11 1
ˆ ˆ( | ) (1 ( ) ( )) ( | 1) (1 ( ) ( )) ( ) ( )
S SS
i i j j i i
ii j i
x k k W k H k x k k W k H k W k z k
== = +
  = − − + −∑∏ ∏     
(18)
For the further information refer to [1].
2.2.3. The Inverse Covariance Method
This method was developed to exploit direct equations for filter response. The matrices which
are inverted are not so massive. State predication and covariance matrices are as in (1) and (2):
1
1 1
1
( | ) ( | 1) ( ) ( ) ( )
S
T
i i i
i
P k k P k k H k R k H k
−
− −
=
 = − + ∑  
    (19)
1 1
1
ˆ ˆ( | ) ( | )[ ( | 1) ( | 1) ( ) ( ) ( )]
S
T
i i i
i
x k k P k k P k k x k k H k R k z k− −
=
= − − + ∑ (20)
2.2.4. The Track-to-track Fusion Method
Track-to-track fusion is an algorithm which combines the estimations which are made at the
place of sensors. Indeed, Kalman Filter estimations are made aside for each sensor.
For each single sensor Kalman filter we have:
[ ]ˆ ˆ ˆ( | ) ( | 1) ( ) ( ) ( ) ( | 1) ,i i i i i ix k k x k k W k z k H k x k k= − + − − (21)
and
( | ) ( | 1) ( ) ( ) ( )T
i i i i iP k k P k k W k S k W k= − − (22)
which
( ) ( ) ( | 1) ( ) ( ) .T
i i i i iS k H k P k k H k R k = − +  (23)
Predictions of local states are made from common states model.
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
5
ˆ ˆ( | 1) ( ) ( 1| 1)i ix k k F k x k k− = − − (24)
and
( | 1) ( ) ( 1 | 1) ( ) ( ) ( ) ( )T T
i iP k k F k P k k F k G k Q k G k− = − − + (25)
So the path fusion algorithm simply computes a weighted average of paths based on variance
weights.
1
1
ˆ ˆ( | ) ( | ) ( | ) ( | )
N
T T i i
i
x k k P k k P k k x k k−
=
= ∑ (26)
1
1
1
( | ) ( | ) .
N
T i
i
P k k P k k
−
−
=
 = ∑  
(27)
This method is not an optimal estimation because of correlation between tracks [8], but it is
simple and functional.
3. PERFORMANCE EVALUATION
3.1 Evaluation Theorize
As stated before, computational complexity of fusion algorithms is among the most important
factors for its implementation. With respect to matrix operations and its complexity especially
for inverse matrix computation, it should be considered as a key factor while hardware or
software implementation. More computational load means more powerful and more expensive
hardware. From another perspective it needs more time to execute computations.
One of the main factors which impact the computational load is the number of sensors or data
fusions sources. This is not necessarily the same for different methods, as there may be a
different method for distinct number of sensors which has the least computational load.
In [1], it is mentioned that for inverse matrix calculation, the computational load is proportional
to the square of the matrix dimensions. Among the proposed methods, the group sensor method
has the most computational complexity. The dimensions of the innovation matrix are
proportional to the number of sensors. This matrix should be inverted in each time steps. As a
result with the increasing number of sensors, the computational load increases more rapidly.
In sequential-sensor method although the innovation matrix dimensions do not change with the
number of sensors, for each sensor an inversing matrix operation has been added. So again the
computational load would be increased, but with a linear rate.
Regardless of the number of sensors employed, the largest matrix inversion that is required is of
dimension the state vector. The addition of new sensors simply requires that the new of
terms 1
( ) ( ) ( )T
i i iH k R k z k−
and 1
( ) ( ) ( )T
i i iH k R k H k−
. Thus the complexity of the update algorithm
grows only linearly with the number of sensors employed. In addition, the update stage can take
place in one single step.
Excellence of the inverse covariance estimator is more obvious when significant number of
sensors is employed. From (10), it is clear that in each cycle of filter, both the prediction
covariance matrix and the updated inverse covariance matrix must be inverted. As a result,
inverse covariance filter shows its superior characteristics just when the dimensions of the
combined observation vector are approximately more than two times of the common state vector
dimensions.
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
6
3.2. Simulation
To proof the claims about the computational load of each method, MATLAB simulation was
applied to them in order to evaluate the required process time of each algorithm. In MATLAB,
“Profile” function allows the user to measure the required process time for each part of the
program.
To increase the accuracy of this function, just one of the cores of the processors in operating
system was used. Also, the priority of MATLAB software was chosen to be in real time state.
To do this, a target trajectory has been modelled. Then, a noise has been added to this trajectory
to model sensors observations. After that, as stated before, Kalman filter multi-sensor data fusion
algorithms has been applied to sensors observations to estimate the target track. Simulation
results are shown in Fig. 1. Note that our emphasis is on computational load of each method.
Also note that just the execution time of the fusion algorithm was measured and path modelling,
noise observations and initial values of parameters processes are not included in the mentioned
time.
Figure1. True state (trajectory), observation and estimation data of 3 sensors
...
Profile on;
...
%codes must be writen here
...
Profile viewer;
...
___ True State (trajectory)
---- Estimated State
... Observations
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
7
4. SIMULATION RESULTS
The modelled system is simulated and the required process times for each algorithm are listed
and shown in TABLE 1 and Fig. 2, respectively.
TABLE 1: The time needed to compute the estimation as a function of increasing number of sensors for
each method.
number
of sensors
1 15 30 45 60 75 90 105 120 150 210 300
group
sensor
0.11 0.31 0.75 1.61 3.01 5 7.64
11.2
5
15.4
9
30.89 75 188
sequential
sensor
0.15 1.41 2.82 4.25 5.68 7.21 8.79
10.4
5
12.1 15.59 23.15
35.6
3
invers
covariance
0.2 0.37 0.56 0.74 0.92 1.1 1.29 1.46 1.66 2.03 2.74 3.86
track to
track
fusion
0.1 1.14 2.71 4.08 5.42 6.76 8.12 9.45 10.8 13.51 18.9
26.9
4
As is seen, in group-sensor method, although a few number of sensors do not need a long time
for data fusion, but, as the number of sensors increases, the computations become more time-
consuming. Sequential sensor method and track-to-track fusion methods need somewhat the
same time. However, these two methods, for the number of sensors greater than 100, are faster
than the group-sensor method. For inverse covariance form the story is somewhat different. For
the number of sensors less than 20, the computational load is approximately the same as the other
methods. However, as the number of sensors increases, the process time is too few in comparison
with the others. This is in agreement with what was said in previous sections.
Figure 2. The required time to compute the estimations. It shows the computational complexity of each
algorithm.
International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013
8
5. CONCLUSION
In order to investigate Multi-sensor single target tracking algorithms base on Kalman filter, four
methods were simulated and compared. For this purpose, the time needed to compute the
estimation as a function of increasing number of sensors for each method is calculated. In
summary, for the number of sensors less than 20, the group-sensor method has the least
computational complexity and the inverse covariance method has the second priority. For greater
number of sensors, inverse covariance method strongly needs the least process time.
Nevertheless, for hardware implementation, type of the hardware (FPGAs, DSPs or so on) and
the mathematical relations of algorithms should be taken into account to choose the best method.
REFERENCES
[1] Hugh Durrant-Whyte, Multi Sensor Data Fusion. Australian Centre for Field Robotics, The
University of Sydney NSW 2006.
[2] Y. Bar-Shalom and X.-R. Li, Multitarget-Multisensor Tracking: Principles and Techniques. ISBN
0-9648312-0-1, 1995.B.D.O. Anderson and J.B. Moore. Optimal Filtering. Prentice Hall, 1979.
[3] Y. Bar-Shalom and T. E. Fortmann, Tracking and Data Association. Academic Press, Orlando,
FL, 1988.
[4] Jitendra R. Raol, Multi-Sensor Data Fusion with MATLAB, CRC Press, Taylor & Francis Group,
2010.
[5] Zou, Wei, and Wei Sun., “ A multi-dimensional data association algorithm for multi-sensor
fusion,” Intelligent Science and Intelligent Data Engineering. Springer Berlin Heidelberg, 2013.
280-288.
[6] Cho, Taehwan, Changho Lee, and Sangbang Choi, “Multi-sensor fusion with interacting multiple
model filter for improved aircraft position accuracy,” Sensors 13.4 (2013): 4122-4137.
[7] Brookner, Eli. Tracking and Kalman filtering made easy. New York: Wiley, 1998.
[8] Y. Bar-Shalom, “On the track to track correlation problem,”. IEEE Trans. Automatic Control,
25(8):802–807, 1981.
Ad

More Related Content

What's hot (18)

Design of State Estimator for a Class of Generalized Chaotic Systems
Design of State Estimator for a Class of Generalized Chaotic SystemsDesign of State Estimator for a Class of Generalized Chaotic Systems
Design of State Estimator for a Class of Generalized Chaotic Systems
ijtsrd
 
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
IJERA Editor
 
IFAC2008art
IFAC2008artIFAC2008art
IFAC2008art
Yuri Kim
 
A comparison of SIFT, PCA-SIFT and SURF
A comparison of SIFT, PCA-SIFT and SURFA comparison of SIFT, PCA-SIFT and SURF
A comparison of SIFT, PCA-SIFT and SURF
CSCJournals
 
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
IJERA Editor
 
I041214752
I041214752I041214752
I041214752
IOSR-JEN
 
The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
theijes
 
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control System
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control SystemChaos Suppression and Stabilization of Generalized Liu Chaotic Control System
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control System
ijtsrd
 
50120130405032
5012013040503250120130405032
50120130405032
IAEME Publication
 
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
csitconf
 
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
cscpconf
 
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
ijait
 
Paper id 26201484
Paper id 26201484Paper id 26201484
Paper id 26201484
IJRAT
 
presentation_btp
presentation_btppresentation_btp
presentation_btp
Anshul Goyal, EIT
 
Analysis of Adaptive Algorithms
Analysis of Adaptive AlgorithmsAnalysis of Adaptive Algorithms
Analysis of Adaptive Algorithms
ijsrd.com
 
Function projective synchronization
Function projective synchronizationFunction projective synchronization
Function projective synchronization
ijcseit
 
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic SystemsA Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
ijtsrd
 
Evaluation of non-parametric identification techniques in second order models...
Evaluation of non-parametric identification techniques in second order models...Evaluation of non-parametric identification techniques in second order models...
Evaluation of non-parametric identification techniques in second order models...
IJECEIAES
 
Design of State Estimator for a Class of Generalized Chaotic Systems
Design of State Estimator for a Class of Generalized Chaotic SystemsDesign of State Estimator for a Class of Generalized Chaotic Systems
Design of State Estimator for a Class of Generalized Chaotic Systems
ijtsrd
 
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
Power System State Estimation Using Weighted Least Squares (WLS) and Regulari...
IJERA Editor
 
IFAC2008art
IFAC2008artIFAC2008art
IFAC2008art
Yuri Kim
 
A comparison of SIFT, PCA-SIFT and SURF
A comparison of SIFT, PCA-SIFT and SURFA comparison of SIFT, PCA-SIFT and SURF
A comparison of SIFT, PCA-SIFT and SURF
CSCJournals
 
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
A Combination of Wavelet Artificial Neural Networks Integrated with Bootstrap...
IJERA Editor
 
I041214752
I041214752I041214752
I041214752
IOSR-JEN
 
The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)The International Journal of Engineering and Science (The IJES)
The International Journal of Engineering and Science (The IJES)
theijes
 
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control System
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control SystemChaos Suppression and Stabilization of Generalized Liu Chaotic Control System
Chaos Suppression and Stabilization of Generalized Liu Chaotic Control System
ijtsrd
 
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
csitconf
 
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
cscpconf
 
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
THE ACTIVE CONTROLLER DESIGN FOR ACHIEVING GENERALIZED PROJECTIVE SYNCHRONIZA...
ijait
 
Paper id 26201484
Paper id 26201484Paper id 26201484
Paper id 26201484
IJRAT
 
Analysis of Adaptive Algorithms
Analysis of Adaptive AlgorithmsAnalysis of Adaptive Algorithms
Analysis of Adaptive Algorithms
ijsrd.com
 
Function projective synchronization
Function projective synchronizationFunction projective synchronization
Function projective synchronization
ijcseit
 
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic SystemsA Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
A Study of Firefly Algorithm and its Application in Non-Linear Dynamic Systems
ijtsrd
 
Evaluation of non-parametric identification techniques in second order models...
Evaluation of non-parametric identification techniques in second order models...Evaluation of non-parametric identification techniques in second order models...
Evaluation of non-parametric identification techniques in second order models...
IJECEIAES
 

Similar to COMPUTATIONAL COMPLEXITY COMPARISON OF MULTI-SENSOR SINGLE TARGET DATA FUSION METHODS BY MATLAB (20)

Refining Underwater Target Localization and Tracking Estimates
Refining Underwater Target Localization and Tracking EstimatesRefining Underwater Target Localization and Tracking Estimates
Refining Underwater Target Localization and Tracking Estimates
CSCJournals
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
IJITCA Journal
 
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
IJITCA Journal
 
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
ijccmsjournal
 
Paper id 26201483
Paper id 26201483Paper id 26201483
Paper id 26201483
IJRAT
 
Design of Kalman filter for Airborne Applications
Design of Kalman filter for Airborne ApplicationsDesign of Kalman filter for Airborne Applications
Design of Kalman filter for Airborne Applications
IJERA Editor
 
presentation.ppt
presentation.pptpresentation.ppt
presentation.ppt
WasiqAli28
 
On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction
cscpconf
 
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTIONON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
cscpconf
 
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
IJERA Editor
 
solver (1)
solver (1)solver (1)
solver (1)
Raj Mitra
 
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
ijistjournal
 
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
ijistjournal
 
Adaptive Control and Synchronization of Hyperchaotic Cai System
Adaptive Control and Synchronization of Hyperchaotic Cai SystemAdaptive Control and Synchronization of Hyperchaotic Cai System
Adaptive Control and Synchronization of Hyperchaotic Cai System
ijctcm
 
Time domain analysis and synthesis using Pth norm filter design
Time domain analysis and synthesis using Pth norm filter designTime domain analysis and synthesis using Pth norm filter design
Time domain analysis and synthesis using Pth norm filter design
CSCJournals
 
Adaptive Projective Lag Synchronization of T and Lu Chaotic Systems
Adaptive Projective Lag Synchronization of T and Lu  Chaotic Systems Adaptive Projective Lag Synchronization of T and Lu  Chaotic Systems
Adaptive Projective Lag Synchronization of T and Lu Chaotic Systems
IJECEIAES
 
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Xin-She Yang
 
Refining Underwater Target Localization and Tracking Estimates
Refining Underwater Target Localization and Tracking EstimatesRefining Underwater Target Localization and Tracking Estimates
Refining Underwater Target Localization and Tracking Estimates
CSCJournals
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTERA NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
A NEW METHOD OF SMALL-SIGNAL CALIBRATION BASED ON KALMAN FILTER
ijcseit
 
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
IJITCA Journal
 
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
APPLYING DYNAMIC MODEL FOR MULTIPLE MANOEUVRING TARGET TRACKING USING PARTICL...
IJITCA Journal
 
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
DYNAMICS, ADAPTIVE CONTROL AND EXTENDED SYNCHRONIZATION OF HYPERCHAOTIC SYSTE...
ijccmsjournal
 
Paper id 26201483
Paper id 26201483Paper id 26201483
Paper id 26201483
IJRAT
 
Design of Kalman filter for Airborne Applications
Design of Kalman filter for Airborne ApplicationsDesign of Kalman filter for Airborne Applications
Design of Kalman filter for Airborne Applications
IJERA Editor
 
presentation.ppt
presentation.pptpresentation.ppt
presentation.ppt
WasiqAli28
 
On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction
cscpconf
 
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTIONON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
cscpconf
 
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
Discretizing of linear systems with time-delay Using method of Euler’s and Tu...
IJERA Editor
 
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
ijistjournal
 
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
GLOBAL CHAOS SYNCHRONIZATION OF UNCERTAIN LORENZ-STENFLO AND QI 4-D CHAOTIC S...
ijistjournal
 
Adaptive Control and Synchronization of Hyperchaotic Cai System
Adaptive Control and Synchronization of Hyperchaotic Cai SystemAdaptive Control and Synchronization of Hyperchaotic Cai System
Adaptive Control and Synchronization of Hyperchaotic Cai System
ijctcm
 
Time domain analysis and synthesis using Pth norm filter design
Time domain analysis and synthesis using Pth norm filter designTime domain analysis and synthesis using Pth norm filter design
Time domain analysis and synthesis using Pth norm filter design
CSCJournals
 
Adaptive Projective Lag Synchronization of T and Lu Chaotic Systems
Adaptive Projective Lag Synchronization of T and Lu  Chaotic Systems Adaptive Projective Lag Synchronization of T and Lu  Chaotic Systems
Adaptive Projective Lag Synchronization of T and Lu Chaotic Systems
IJECEIAES
 
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Applications and Analysis of Bio-Inspired Eagle Strategy for Engineering Opti...
Xin-She Yang
 
Ad

Recently uploaded (20)

Compiler Design_Syntax Directed Translation.pptx
Compiler Design_Syntax Directed Translation.pptxCompiler Design_Syntax Directed Translation.pptx
Compiler Design_Syntax Directed Translation.pptx
RushaliDeshmukh2
 
W1 WDM_Principle and basics to know.pptx
W1 WDM_Principle and basics to know.pptxW1 WDM_Principle and basics to know.pptx
W1 WDM_Principle and basics to know.pptx
muhhxx51
 
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E..."Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
Infopitaara
 
Development of MLR, ANN and ANFIS Models for Estimation of PCUs at Different ...
Development of MLR, ANN and ANFIS Models for Estimation of PCUs at Different ...Development of MLR, ANN and ANFIS Models for Estimation of PCUs at Different ...
Development of MLR, ANN and ANFIS Models for Estimation of PCUs at Different ...
Journal of Soft Computing in Civil Engineering
 
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
charlesdick1345
 
2025 Apply BTech CEC .docx
2025 Apply BTech CEC                 .docx2025 Apply BTech CEC                 .docx
2025 Apply BTech CEC .docx
tusharmanagementquot
 
some basics electrical and electronics knowledge
some basics electrical and electronics knowledgesome basics electrical and electronics knowledge
some basics electrical and electronics knowledge
nguyentrungdo88
 
How to use nRF24L01 module with Arduino
How to use nRF24L01 module with ArduinoHow to use nRF24L01 module with Arduino
How to use nRF24L01 module with Arduino
CircuitDigest
 
Structural Response of Reinforced Self-Compacting Concrete Deep Beam Using Fi...
Structural Response of Reinforced Self-Compacting Concrete Deep Beam Using Fi...Structural Response of Reinforced Self-Compacting Concrete Deep Beam Using Fi...
Structural Response of Reinforced Self-Compacting Concrete Deep Beam Using Fi...
Journal of Soft Computing in Civil Engineering
 
Routing Riverdale - A New Bus Connection
Routing Riverdale - A New Bus ConnectionRouting Riverdale - A New Bus Connection
Routing Riverdale - A New Bus Connection
jzb7232
 
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
SiddhiKulkarni18
 
Data Structures_Introduction to algorithms.pptx
Data Structures_Introduction to algorithms.pptxData Structures_Introduction to algorithms.pptx
Data Structures_Introduction to algorithms.pptx
RushaliDeshmukh2
 
Process Parameter Optimization for Minimizing Springback in Cold Drawing Proc...
Process Parameter Optimization for Minimizing Springback in Cold Drawing Proc...Process Parameter Optimization for Minimizing Springback in Cold Drawing Proc...
Process Parameter Optimization for Minimizing Springback in Cold Drawing Proc...
Journal of Soft Computing in Civil Engineering
 
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
Taqyea
 
Compiler Design Unit1 PPT Phases of Compiler.pptx
Compiler Design Unit1 PPT Phases of Compiler.pptxCompiler Design Unit1 PPT Phases of Compiler.pptx
Compiler Design Unit1 PPT Phases of Compiler.pptx
RushaliDeshmukh2
 
MODULE 03 - CLOUD COMPUTING- [BIS 613D] 2022 scheme.pptx
MODULE 03 - CLOUD COMPUTING-  [BIS 613D] 2022 scheme.pptxMODULE 03 - CLOUD COMPUTING-  [BIS 613D] 2022 scheme.pptx
MODULE 03 - CLOUD COMPUTING- [BIS 613D] 2022 scheme.pptx
Alvas Institute of Engineering and technology, Moodabidri
 
ISO 9001 quality management systemPPT.pptx
ISO 9001 quality management systemPPT.pptxISO 9001 quality management systemPPT.pptx
ISO 9001 quality management systemPPT.pptx
mesfin608
 
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptxExplainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
MahaveerVPandit
 
Data Structures_Linear data structures Linked Lists.pptx
Data Structures_Linear data structures Linked Lists.pptxData Structures_Linear data structures Linked Lists.pptx
Data Structures_Linear data structures Linked Lists.pptx
RushaliDeshmukh2
 
AI-assisted Software Testing (3-hours tutorial)
AI-assisted Software Testing (3-hours tutorial)AI-assisted Software Testing (3-hours tutorial)
AI-assisted Software Testing (3-hours tutorial)
Vəhid Gəruslu
 
Compiler Design_Syntax Directed Translation.pptx
Compiler Design_Syntax Directed Translation.pptxCompiler Design_Syntax Directed Translation.pptx
Compiler Design_Syntax Directed Translation.pptx
RushaliDeshmukh2
 
W1 WDM_Principle and basics to know.pptx
W1 WDM_Principle and basics to know.pptxW1 WDM_Principle and basics to know.pptx
W1 WDM_Principle and basics to know.pptx
muhhxx51
 
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E..."Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
"Boiler Feed Pump (BFP): Working, Applications, Advantages, and Limitations E...
Infopitaara
 
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
DATA-DRIVEN SHOULDER INVERSE KINEMATICS YoungBeom Kim1 , Byung-Ha Park1 , Kwa...
charlesdick1345
 
some basics electrical and electronics knowledge
some basics electrical and electronics knowledgesome basics electrical and electronics knowledge
some basics electrical and electronics knowledge
nguyentrungdo88
 
How to use nRF24L01 module with Arduino
How to use nRF24L01 module with ArduinoHow to use nRF24L01 module with Arduino
How to use nRF24L01 module with Arduino
CircuitDigest
 
Routing Riverdale - A New Bus Connection
Routing Riverdale - A New Bus ConnectionRouting Riverdale - A New Bus Connection
Routing Riverdale - A New Bus Connection
jzb7232
 
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
Comparing Emerging Cloud Platforms –like AWS, Azure, Google Cloud, Cloud Stac...
SiddhiKulkarni18
 
Data Structures_Introduction to algorithms.pptx
Data Structures_Introduction to algorithms.pptxData Structures_Introduction to algorithms.pptx
Data Structures_Introduction to algorithms.pptx
RushaliDeshmukh2
 
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
最新版加拿大魁北克大学蒙特利尔分校毕业证(UQAM毕业证书)原版定制
Taqyea
 
Compiler Design Unit1 PPT Phases of Compiler.pptx
Compiler Design Unit1 PPT Phases of Compiler.pptxCompiler Design Unit1 PPT Phases of Compiler.pptx
Compiler Design Unit1 PPT Phases of Compiler.pptx
RushaliDeshmukh2
 
ISO 9001 quality management systemPPT.pptx
ISO 9001 quality management systemPPT.pptxISO 9001 quality management systemPPT.pptx
ISO 9001 quality management systemPPT.pptx
mesfin608
 
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptxExplainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
Explainable-Artificial-Intelligence-XAI-A-Deep-Dive (1).pptx
MahaveerVPandit
 
Data Structures_Linear data structures Linked Lists.pptx
Data Structures_Linear data structures Linked Lists.pptxData Structures_Linear data structures Linked Lists.pptx
Data Structures_Linear data structures Linked Lists.pptx
RushaliDeshmukh2
 
AI-assisted Software Testing (3-hours tutorial)
AI-assisted Software Testing (3-hours tutorial)AI-assisted Software Testing (3-hours tutorial)
AI-assisted Software Testing (3-hours tutorial)
Vəhid Gəruslu
 
Ad

COMPUTATIONAL COMPLEXITY COMPARISON OF MULTI-SENSOR SINGLE TARGET DATA FUSION METHODS BY MATLAB

  • 1. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 DOI : 10.5121/ijccms.2013.2201 1 COMPUTATIONAL COMPLEXITY COMPARISON OF MULTI-SENSOR SINGLE TARGET DATA FUSION METHODS BY MATLAB Sayed Amir Hoseini1 and Mohammad Reza Ashraf 2 1 Department of Electrical Engineering, Amirkabir University of technology, Tehran, Iran [email protected] 2 Department of Electrical Engineering, University of Tehran, Tehran, Iran [email protected] ABSTRACT Target tracking using observations from multiple sensors can achieve better estimation performance than a single sensor. The most famous estimation tool in target tracking is Kalman filter. There are several mathematical approaches to combine the observations of multiple sensors by use of Kalman filter. An important issue in applying a proper approach is computational complexity. In this paper, four data fusion algorithms based on Kalman filter are considered including three centralized and one decentralized methods. Using MATLAB, computational loads of these methods are compared while number of sensors increases. The results show that inverse covariance method has the best computational performance if the number of sensors is above 20. For a smaller number of sensors, other methods, especially group sensors, are more appropriate.. KEYWORDS Data fusion, Target Tracking, Kalman Filter, Multi-sensor, MATLAB 1. INTRODUCTION Data fusion is the process of combining information from a number of different sources to provide a robust and complete description of an environment or process of interest. Data fusion is of special significance in any application where large amounts of data must be combined, fused and distilled to obtain information of appropriate quality and integrity on which decisions can be made. Data fusion finds application in many military systems, in civilian surveillance and monitoring tasks, in process control and in information systems. Data fusion methods are particularly important in the drive toward autonomous systems in all these applications [1]. Estimation is the single most important problem in sensor data fusion. Fundamentally, an estimator is a decision rule which takes as an argument a sequence of observations and whose action is to compute a value for the parameter or state of interest. Almost all data fusion problems involve this estimation process: we obtain a number of observations from a group of sensors and using this information we wish to find some estimate of the true state of the environment we are observing. Estimation encompasses all important aspects of the data fusion problem. The most famous estimation tool in target tracking is Kalman filter. Tracking targets with Kalman filtering is an active research area and there are substantial literatures in this field such as [2], [3] and [4]. [1,4] talk about several multi-sensor data fusion methods. [5] and [6] are also the same but latest attempts. In [1], the computational complexities of these methods are compared based on mathematical formulation, but no simulation or implementation proof is presented. In this paper, the aim is to continue the related work in [1] and evaluate its claim using MATLAB simulations.
  • 2. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 2 This paper begins with a brief summary of the Kalman filter algorithm. The intention is to introduce notation and key data fusion concepts; Prior familiarity with the basic Kalman Filter algorithm is assumed. The multi-sensor Kalman filter is then discussed. Four main algorithms are considered; the group-sensor method, the sequential sensor method, the inverse covariance form and the track-to-track fusion. For the related concepts and additional formulation refer to [1]. After that, in section 3, these methods are evaluated in theory and simulation. Finally, Simulation results and a comparison of these methods are presented in section 4. 2. KALMAN FILTER IN TARGET TRACKING The Kalman Filter is a recursive linear estimator which successively calculates an estimate for a continuous valued state, that evolves over time, on the basis of periodic observations that of this state. 2.1. State and Observation Models The starting point for the Kalman filter algorithm is to define a model for the states to be estimated in the standard state-space form and observation model for data received from sensor [7]. A general motion model used in discrete Kalman filter for target tracking is: ( ) ( ) ( 1) ( ) ( )x k F k x k G k v k= − + (1) ( ) ( ) ( ) ( )z k H k x k w k= + (2) where F(k) is the state transition matrix, x(k) is the state vector at time k, G(k) is the process noise gain matrix, The process noise v(k) and the measurement noise w(k) are zero mean, mutually independent, white, Gaussian with covariance Q and R respectively. z(k) is the measurement vector at time k and H(k) is observation matrix of the states computed at time k. { } { }( ) ( ) 0, ,E v k E w k k= = ∀ (3) { } { }( ) ( ) ( ), ( ) ( ) ( ).T T ij ijE v i v j Q i E w i w j R i = = (4) 2.1.1. State Prediction The state and state covariance matrix at time k-1 are predicted to time k as follows: ˆ ˆ( | 1) ( ) ( 1| 1)x k k F k x k k− = − − (5) ( | 1) ( ) ( 1| 1) ( ) ( ) ( ) ( ),T T P k k F k P k k F k G k Q k G k− = − − + (6) where ˆ( 1| 1)x k k− − is the estimated state vector at time k, ( 1| 1)P k k− − is the estimated state covariance matrix at the same time, ˆ( | 1)x k k − is the predicted state and ( | 1)P k k − is the predicted state covariance matrix. 2.1.2. Measurement update At time k an observation ( )z k is made and the updated estimate ˆ( | )x k k of the state ( )x k , together with the updated estimate covariance ( )|P k k is computed from the state prediction and observation according to: ˆ ˆ ˆ( | ) ( | 1) ( )( ( ) ( ) ( | 1))x k k x k k W k z k H k x k k= − + − − (7) and ( | ) (1 ( ) ( )) ( | 1)(1 ( ) ( )) ( ) ( ) ( )T T P k k W k H k P k k W k H k W k R k W k= − − − + (8) where the gain matrix ( )W k is given by:
  • 3. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 3 1 ( ) ( | 1) ( ) ( ) ( | 1) ( ) ( )T W k P k k H k H k P k k H k R k −  = − − +  (9) 2.2. The Multi-Sensor Kalman Filter Many of the techniques developed for single sensor Kalman filters can be applied directly to multi-sensor estimation and tracking problems. In principle, a group of sensors can be considered as a single sensor with a large and possibly complex observation model. In this case the Kalman filter algorithm is directly applicable to the multi-sensor estimation problem. This technique is called “group-sensor method”. However, as will be seen, this approach is practically limited to relatively small numbers of sensors. A second approach is to consider each observation made by each sensor as a separate and independent realization, made according to a specific observation model, which can be incorporate into the estimate in a sequential manner. Again, single-sensor estimation techniques, applied sequentially, can be applied to this formulation of the multi-sensor estimation problem. This technique is called “sequential-sensor method”. However, as will be seen, this approach requires that a new prediction and gain matrix be calculated for each observation from each sensor at every time-step, and so is computationally very expensive. A third approach is to explicitly derive equations for integrating multiple observations made at the same time into a common state estimate. Starting from the formulation of the multi-sensor Kalman filter algorithm, employing a single model for a group of sensors, a set of recursive equations for integrating individual sensor observations can be derived. This method is called “inverse covariance form”. The systems considered to this point are all ‘centralized’; the observations made by sensors are reported back to a central processing unit in a raw form where they are processed by a single algorithm in much the same way as single sensor systems. It is also possible to formulate the multi-sensor estimation problem in terms of a number of local sensor filters, each generating state estimates, which are subsequently communicated in processed form back to a central fusion center. This distributed processing structure has a number of advantages in terms of modularity of the resulting architecture. However, the algorithms required to fuse estimate or track information at the central site can be quite complex. This Fusion method is called “track-to-track fusion” and discus as a fourth approach. In the future equations, Indexes indicate the corresponding sensor number. For example, ( )iz k is the observation of ith sensor at the kth time step. 2.2.1. The Group-Sensor Method The simplest way of dealing with a multi-sensor estimation problem is to combine all observations and observation models in to a single composite ‘group sensor’ and then to deal with the estimation problem using an identical algorithm to that employed in single-sensor systems. The combinatorial observation vector is defined as: 1( ) ( ),..., ( ) , TT T sz k z k z k   (10) and combinatorial observation model is as: 1( ) ( ),..., ( ) TT T sH k H k H k   (11) and 1( ) ( ),..., ( ) , TT T sw k w k w k   (12)
  • 4. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 4 and { } { } { } 1 1 1 ( ) ( ) ( ) ( ),..., ( ) ( ),..., ( ) ( ),..., ( ) , TT T T T T s s s R k E w k w k w k w k w k w k blockdiag R k R k    = = Ε =    = (13) Observation noise covariance is as a block-diagonal matrix which each block of diagonal is the observation noise matrix of each sensor. Observation vector are made in a combinatorial manner. Predication equations are like single sensor Kalman filter (equations (1) and (2)). 2.2.2 The Sequential-Sensor Method A second approach to the multi-sensor estimation problem is to consider each sensor observation as an independent, sequential update to the state estimate and for each observation to compute an appropriate prediction and gain matrix. The final formulation provide below: ˆ ˆ( | 1) ( ) ( 1| 1)x k k F k x k k− = − − (14) ( | 1) ( ) ( 1| 1) ( ) ( ) ( ) ( ).T T P k k F k P k k F k G k Q k G k− = − − + (15) ( ) ( ) ( | , 1) ( ) ( )T p p p pS k H k P k k p H k R k= − + (16) 1 ( ) ( | , 1) ( ) ( )T p p pW k P k k p H k S k− = − (17) 11 1 ˆ ˆ( | ) (1 ( ) ( )) ( | 1) (1 ( ) ( )) ( ) ( ) S SS i i j j i i ii j i x k k W k H k x k k W k H k W k z k == = +   = − − + −∑∏ ∏      (18) For the further information refer to [1]. 2.2.3. The Inverse Covariance Method This method was developed to exploit direct equations for filter response. The matrices which are inverted are not so massive. State predication and covariance matrices are as in (1) and (2): 1 1 1 1 ( | ) ( | 1) ( ) ( ) ( ) S T i i i i P k k P k k H k R k H k − − − =  = − + ∑       (19) 1 1 1 ˆ ˆ( | ) ( | )[ ( | 1) ( | 1) ( ) ( ) ( )] S T i i i i x k k P k k P k k x k k H k R k z k− − = = − − + ∑ (20) 2.2.4. The Track-to-track Fusion Method Track-to-track fusion is an algorithm which combines the estimations which are made at the place of sensors. Indeed, Kalman Filter estimations are made aside for each sensor. For each single sensor Kalman filter we have: [ ]ˆ ˆ ˆ( | ) ( | 1) ( ) ( ) ( ) ( | 1) ,i i i i i ix k k x k k W k z k H k x k k= − + − − (21) and ( | ) ( | 1) ( ) ( ) ( )T i i i i iP k k P k k W k S k W k= − − (22) which ( ) ( ) ( | 1) ( ) ( ) .T i i i i iS k H k P k k H k R k = − +  (23) Predictions of local states are made from common states model.
  • 5. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 5 ˆ ˆ( | 1) ( ) ( 1| 1)i ix k k F k x k k− = − − (24) and ( | 1) ( ) ( 1 | 1) ( ) ( ) ( ) ( )T T i iP k k F k P k k F k G k Q k G k− = − − + (25) So the path fusion algorithm simply computes a weighted average of paths based on variance weights. 1 1 ˆ ˆ( | ) ( | ) ( | ) ( | ) N T T i i i x k k P k k P k k x k k− = = ∑ (26) 1 1 1 ( | ) ( | ) . N T i i P k k P k k − − =  = ∑   (27) This method is not an optimal estimation because of correlation between tracks [8], but it is simple and functional. 3. PERFORMANCE EVALUATION 3.1 Evaluation Theorize As stated before, computational complexity of fusion algorithms is among the most important factors for its implementation. With respect to matrix operations and its complexity especially for inverse matrix computation, it should be considered as a key factor while hardware or software implementation. More computational load means more powerful and more expensive hardware. From another perspective it needs more time to execute computations. One of the main factors which impact the computational load is the number of sensors or data fusions sources. This is not necessarily the same for different methods, as there may be a different method for distinct number of sensors which has the least computational load. In [1], it is mentioned that for inverse matrix calculation, the computational load is proportional to the square of the matrix dimensions. Among the proposed methods, the group sensor method has the most computational complexity. The dimensions of the innovation matrix are proportional to the number of sensors. This matrix should be inverted in each time steps. As a result with the increasing number of sensors, the computational load increases more rapidly. In sequential-sensor method although the innovation matrix dimensions do not change with the number of sensors, for each sensor an inversing matrix operation has been added. So again the computational load would be increased, but with a linear rate. Regardless of the number of sensors employed, the largest matrix inversion that is required is of dimension the state vector. The addition of new sensors simply requires that the new of terms 1 ( ) ( ) ( )T i i iH k R k z k− and 1 ( ) ( ) ( )T i i iH k R k H k− . Thus the complexity of the update algorithm grows only linearly with the number of sensors employed. In addition, the update stage can take place in one single step. Excellence of the inverse covariance estimator is more obvious when significant number of sensors is employed. From (10), it is clear that in each cycle of filter, both the prediction covariance matrix and the updated inverse covariance matrix must be inverted. As a result, inverse covariance filter shows its superior characteristics just when the dimensions of the combined observation vector are approximately more than two times of the common state vector dimensions.
  • 6. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 6 3.2. Simulation To proof the claims about the computational load of each method, MATLAB simulation was applied to them in order to evaluate the required process time of each algorithm. In MATLAB, “Profile” function allows the user to measure the required process time for each part of the program. To increase the accuracy of this function, just one of the cores of the processors in operating system was used. Also, the priority of MATLAB software was chosen to be in real time state. To do this, a target trajectory has been modelled. Then, a noise has been added to this trajectory to model sensors observations. After that, as stated before, Kalman filter multi-sensor data fusion algorithms has been applied to sensors observations to estimate the target track. Simulation results are shown in Fig. 1. Note that our emphasis is on computational load of each method. Also note that just the execution time of the fusion algorithm was measured and path modelling, noise observations and initial values of parameters processes are not included in the mentioned time. Figure1. True state (trajectory), observation and estimation data of 3 sensors ... Profile on; ... %codes must be writen here ... Profile viewer; ... ___ True State (trajectory) ---- Estimated State ... Observations
  • 7. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 7 4. SIMULATION RESULTS The modelled system is simulated and the required process times for each algorithm are listed and shown in TABLE 1 and Fig. 2, respectively. TABLE 1: The time needed to compute the estimation as a function of increasing number of sensors for each method. number of sensors 1 15 30 45 60 75 90 105 120 150 210 300 group sensor 0.11 0.31 0.75 1.61 3.01 5 7.64 11.2 5 15.4 9 30.89 75 188 sequential sensor 0.15 1.41 2.82 4.25 5.68 7.21 8.79 10.4 5 12.1 15.59 23.15 35.6 3 invers covariance 0.2 0.37 0.56 0.74 0.92 1.1 1.29 1.46 1.66 2.03 2.74 3.86 track to track fusion 0.1 1.14 2.71 4.08 5.42 6.76 8.12 9.45 10.8 13.51 18.9 26.9 4 As is seen, in group-sensor method, although a few number of sensors do not need a long time for data fusion, but, as the number of sensors increases, the computations become more time- consuming. Sequential sensor method and track-to-track fusion methods need somewhat the same time. However, these two methods, for the number of sensors greater than 100, are faster than the group-sensor method. For inverse covariance form the story is somewhat different. For the number of sensors less than 20, the computational load is approximately the same as the other methods. However, as the number of sensors increases, the process time is too few in comparison with the others. This is in agreement with what was said in previous sections. Figure 2. The required time to compute the estimations. It shows the computational complexity of each algorithm.
  • 8. International Journal of Chaos, Control, Modelling and Simulation (IJCCMS) Vol.2, No.2, June 2013 8 5. CONCLUSION In order to investigate Multi-sensor single target tracking algorithms base on Kalman filter, four methods were simulated and compared. For this purpose, the time needed to compute the estimation as a function of increasing number of sensors for each method is calculated. In summary, for the number of sensors less than 20, the group-sensor method has the least computational complexity and the inverse covariance method has the second priority. For greater number of sensors, inverse covariance method strongly needs the least process time. Nevertheless, for hardware implementation, type of the hardware (FPGAs, DSPs or so on) and the mathematical relations of algorithms should be taken into account to choose the best method. REFERENCES [1] Hugh Durrant-Whyte, Multi Sensor Data Fusion. Australian Centre for Field Robotics, The University of Sydney NSW 2006. [2] Y. Bar-Shalom and X.-R. Li, Multitarget-Multisensor Tracking: Principles and Techniques. ISBN 0-9648312-0-1, 1995.B.D.O. Anderson and J.B. Moore. Optimal Filtering. Prentice Hall, 1979. [3] Y. Bar-Shalom and T. E. Fortmann, Tracking and Data Association. Academic Press, Orlando, FL, 1988. [4] Jitendra R. Raol, Multi-Sensor Data Fusion with MATLAB, CRC Press, Taylor & Francis Group, 2010. [5] Zou, Wei, and Wei Sun., “ A multi-dimensional data association algorithm for multi-sensor fusion,” Intelligent Science and Intelligent Data Engineering. Springer Berlin Heidelberg, 2013. 280-288. [6] Cho, Taehwan, Changho Lee, and Sangbang Choi, “Multi-sensor fusion with interacting multiple model filter for improved aircraft position accuracy,” Sensors 13.4 (2013): 4122-4137. [7] Brookner, Eli. Tracking and Kalman filtering made easy. New York: Wiley, 1998. [8] Y. Bar-Shalom, “On the track to track correlation problem,”. IEEE Trans. Automatic Control, 25(8):802–807, 1981.