Radar and AIS Data Fusion for the Needs of the Maritime

Navigation

Andrzej Stateczny, Andrzej Lisaj

Maritime University of Szczecin

Wały Chrobrego 1-2, 70-500 Szczecin, Poland
e-mail: astateczny@am.szczecin.pl, alisaj@am.szczecin.pl

Abstract: Integration the following components: RADAR and AIS, offer the
opportunity to build more sophisticated systems for safety navigation. The
integration’s idea is set on the exploitation of synergies between both components.
The paper presents results of development of method of integration of radar data
and AIS information on the base of multisensor Kalman filter.

1. Introduction
Ships have two sources for the display of own and other ships positions: Radar and its
complement: AIS. Following simple logic, if one system is the complement to the other, they
should have similar and compatible results. AIS-positions should "complement" radar echoes.
The display error between two moving objects is dependent upon the positioning accuracy of
each object plus the vector of relative horizontal translation of these objects during the up-
date interval, as a function of their relative speed. To match AIS and ARPA-targets on the
display of such a device, AIS-data do not need to be updated within the same intervals as
radar. It is assumed that once marked with the AIS-data, they will follow the ARPA-target at
any interval. If the transmitted AIS position should be fused with the (filtered) radar echo, not
only the update rate < radar update (i.e. 3 sec) must be secured, but also clear regulations and
standardized (verifiable) algorithms for the data fusing process.
Situation to have different track of the same object is unacceptable for watching
officer or VTS operator. AIS and radar tracks should not be displayed separately. The
integration radar and transponder information becomes a typical problem of track fusion. Both
information sources should be used to estimation of target position and movement. One of the
solutions is to use multisensor Kalman filter.

2. Multisensor Kalman filter

AIS and radar data fusion consist in:
^ ^ ^
X 0 (t \ t ) = A1(t ) X 1(t \ t ) + ..... + Al (t ) X l (t \ t ) (1)
where:
- A i (t ), i = 1,2,...., l , are weights matrixes,
^
- X 0 (t \ t ) are estimation values from particular date sets with 1,2,.......l , sensors directly
consequential from the filtration process.
Weights matrix is calculate from equation:

A = Σ −1e(e Τ Σ −1e) −1 (2)

where :
- A = [ A1, A2,...... Al ]Τ , matrix A consist of blocks A1, A2,...... Al where each of them has n x n
dimension where n is variables amount entering to the object state vector:
- e = [ I 1 ,.....I l ]Τ , matrix e consist of blocks I1 ,.....I l where each block I is single matrix by
mainly diagonal at n x n dimension.
⎡11 0 ⎤
⎢0 1 ⎥
⎢ 1 ⎥
⎡ A1 ⎤
⎢"""⎥ ⎢ ⎥
⎢ ⎥ #
e = ⎢% ⎥ nxl A = ⎢ ⎥ nxl
⎢# ⎥
⎢"""⎥ ⎢ ⎥
⎢ ⎥ ⎣⎢ A l ⎦⎥
⎢1n 0 ⎥
⎢0 1 ⎥
⎣ n ⎦

Matrix Σ is matrix diagonal compound from diagonal elements in covariance matrix Pij.

Σ = ( Pij )nlxnl where P ij , i, j = 1,2,...l (3)

⎡ P11 P12 .... .... .... P1n ⎤

⎢P P22 .... .... .... P2 n ⎥⎥
⎡ Pij .... Pij ⎤ ⎢ 21
⎢ .... .... .... .... .... ... ⎥
Σ = ⎢⎢ ... ... ... ⎥⎥ Pij = ⎢ ⎥
⎢⎣ Pij ... Pij ⎥⎦ lxl ⎢ .... ..... .... .... .... ... ⎥
⎢ .... ..... ..... .... .... ..... ⎥
⎢ ⎥
⎢⎣ Pn1 Pn 2 ..... .... .... Pnn ⎥⎦ nxn

Covariance matrix is calculate from formula:

Pij (t + 1 \ t + 1) = [ I n − K i (t + 1) H i ] * [ΦPij (t \ t )Φ Τ + ΓQΓ Τ ] * [ I n − K j (t + 1) H j ]Τ (4)

where : Pij(t\t), i,j = 1,2....l are interaction matrixes covariance of errors after filtrations. If
i≠j are elements outside diagonal mainly in matrix.

Initial values Pij(0\0)=P0,where P0 is evaluation of accuracy date fusion is calculate from

formula:

P0 = (e Τ Σ −1e) −1 (5)
or
P0 = (e Τ Σ −1e) −1 e Τ Σ −1 ( Pij )nlxnlΣ −1e(e Τ Σ −1e) −1 . (6)

Next steps of date fusion process is define local Kalman filters, where they quantity is equal
particular sensors.

Initial values:
^
x ( 0 | 0) = μ 0 , Pi (0 | 0) = P0 (7)
 ⎡σ 12 0 0⎤
⎢ ⎥
P0 =Pij(0\0) = ⎢ 0 σ 2
2 0 ⎥ , exemplary σ 12 = σ 22 = σ 32 = 0.1
⎢0 0 σ 32 ⎥⎦ nxn
⎣

Kalman filter we can describe by means of following equations:

Pi (t + 1 \ t + 1) = [ I n − K i (t + 1) H i ]Pi (t + 1 \ t ) (8)
Pi (t + 1 \ t ) = ΦPi (t \ t )Φ Τ + ΓQΓ Τ (9)
Τ Τ
K i (t + 1) = Pi (t + 1 \ t ) H * [ H i Pi (t + 1 \ t ) H + Ri ]
i i (10)
^
ε i (t + 1) = y i (t + 1) − H i x(t + 1 \ t ) (11)
^ ^
X (t + 1 \ t ) = Φ X (t \ t ) (12)
^ ^
X (t + 1 \ t + 1) = X (t + 1 \ t ) + K i (t + 1)ε i (t + 1) (13)

4. Numerical experiment
Research for the needs of this article was conducted by means of a radar and AIS data.
Radar and AIS data where simulated in different step of time and its own distribution of errors
taken on the basis of real measurements. Results of data fusion was presented on fig. 1.

Simulated trajectory
Measurements
2400 Results

2350

2300
Distance[m]

2250

2200

2150

2100

2050

-150 -100 -50 0 50 100 150 200 250 300 350

t[s]

Fig. 1. The results of numerical experiment of data fusion for simulated scenario.
5. Conclusions
Thanks radar and AIS data fusion, the watching officer would receive more complete
navigational information, even about objects untracked by its own ship’s ARPA.
The most important reason is that radar can obtain image including each targets also
unequipped AIS receivers. The main advantage of AIS is that it allows navigators to “look
around the corner” in narrow passages which means that they can also recognize vessels with
due to insufficient radar coverage would normally be not detect.

Reference:
[1] Arsenault R and others, Fusing Information in 3D Chart-of-the-Future Display. Proceedings
of the 11th IAIN World Congress “Smart navigation – Systems and Services”, Berlin 2003.
[2] Frerichs W., Vision 2002: Integrated Bridge Systems – looking ahead. European Journal of
Navigation, vol. 1, No 1, 2003.
[3] Hagiwara H., New Traffic Management System based on AIS and Planned Route
Information. Proceedings of the 11th IAIN World Congress “Smart navigation – Systems and
Services”, Berlin 2003.
[4] Harre I., AIS and VTS. European Journal of Navigation, vol. 1, No 1, 2003.
[5] Lisaj A, Mohammad C.: Intelligent Fusion of Navigational Data, Polish Journal of
Environmental Studies Vol. 14, Supplement I, 2005.
[6] Pettersson B., The Implication of AIS –Automatic Identification System. European Journal of
Navigation, vol. 1, No 2, 2003.
[7] Sandler M., Gern T., Zimmermann R., Integration of Inland ECDIS, Radar and AIS. European
Journal of Navigation, vol. 1, No 2, 2003.
[8] Shu-li Sun.: Multi-sensor optimal information fusion Kalman filters with application,
Aerospace Science and Technology 2003.
[9] Stateczny A., AIS and RADAR Data Fusion for Maritime Navigation. ZN AM, Szczecin
2004.
[10] Stateczny A., Comparative Navigation as an Alternative Positioning System. Proceedings of
the 11th IAIN World Congress “Smart navigation – Systems and Services”, Berlin 2003.
[11] Stateczny A., Integration of RADAR, EGNOS/GALILEO, AIS and 3D ECDIS. International
Radar Symposium, Warszawa, 2004.
[12] Stateczny A., Multisensor Navigational Data Fusion in an Active Vessel Traffic System.
Proceedings of International Conference on Marine Navigation and Technology “MELAHA
2004” organized by Arab Institute of Navigation Cairo Egypt 2004.
[13] Stateczny A., The Concept of Active Vessel Traffic Management and Information System.
Proceedings of the 7th Poland Italy Geodetic Meeting Bresanone 2003. Reports on Geodesy
No.2(65), 2003.