|
You are at: Wagner
Home > Technologies > Data Fusion Tracking > Stick-Figure Tracker
Airport Ground
Tracking Radar Stick-Figure Tracker
This page describes a new tracking algorithm, called the Stick-Figure
tracker, and the implementation of this tracker in prototype
software.
Algorithm patented by,
Joseph H. Discenza and C. Allen Butler,
Stick Figure Radar Tracking Process
Patent No. 5,300,933.
Contents
The
Ground Traffic Radar Problem
| The ASDE-3 radar is a special purpose, high-resolution
surface radar used to monitor the movement of airport ground traffic.
Because of the ratio of target size to radar range, a jetliner can return
energy in a very large number of range and bearing sample bins. Moreover,
because of differences in reflectivity, the target may actually appear in
the data as two or more smaller returns rather than one large one. In some
cases, the radar picture may look very much like the plan view of the
aircraft itself, while, in others it may not. |
Depiction of radar return from a large jetliner on a runway. |
The current method of tracking in the Airport Movement Area
Safety System (AMASS) program involves detecting a region of
greater radar return in the (range, bearing) of the radar signal
and then computing a centroid from all the reflected energy of
the target. That centroid location is then tracked using an a-b
tracker (see [4], chapter 2). Because
different components of the aircraft can reflect energy differently
depending on aspect, the radar can miss portions of the aircraft
on any one sweep (see [3]).
This can cause the centroid to exhibit significant jumps, which
tends to confuse the tracker, especially during aircraft turns.
This phenomena might result in the erroneous generation of an
alert if the tracker output were being used to predict runway
incursions. Such false alarms tend limit the usefulness of the
system and degrade the confidence of the user. Targets such as
trucks, buses, and smaller aircraft that have smaller radar cross-sections
do not exhibit the kinds of tracking anomalies produced by large
jets.
The
Stick-Figure Tracking Approach
The idea behind the Stick-Figure tracker is that individual
elements of the radar returns of a large jetliner can be distinguished,
depending on the configuration of the individual aircraft type
and radar aspect angle. Examples of individual elements, or components,
may be the nose, tail, wings, and fuselage. If the data extraction
process can be modified to measure the range and bearing of each
component, then each component can be tracked individually. A
separate process can associate the component tracks together,
estimate the true center and orientation of the aircraft and
predict the precise locations of critical physical components
(e.g., wingtips, nose, tail). In order to accomplish this, the
composite tracker needs a geometric definition of the relationships
of the radar components to one another and to the geometry of
the aircraft itself. The defined geometry is called the "Stick-Figure".
Typical five-component jetliner stick-figure geometry model. |
Diagram of the Component Capture Windows for the DC-9 Sample Data Set. |
If a set of component measurements (radar centroids) are observed
then the least squares estimate of the center and the orientation
angle of the aircraft can be easily computed. The details of
the computation are included in the patent document. They are
based on standard multiple regression techniques.
The first step in using this technique is to determine which
portions of the radar energy belong to which components. We examined
the return profiles of a DC-9 in sample radar data provided by
the FAA Then we created five rectangular regions corresponding
to these energy distributions.
These windows are then used to partition the radar returns
from the aircraft into individual components.
Using the windows obtained from above we calculated a set of
component centroids from nine scans of data. The average of these
gives us a stick figure for this aircraft and aspect angle, depicted
below superimposed on a DC-9 diagram.
Stick Figure geometry of a DC-9, superimposed on aircraft's plan
view.
Analysis
of Results
In order to demonstrate the utility of the Stick-Figure tracker,
we chose three different scenarios from the data sets provided
by the government. In one of the scenarios, the window includes
the take-off area for runway 28C. It is centered approximately
2,347 meters at 104 degrees from the control tower
and covers an area of approximately 385 meters (in azimuth)
and 378 meters (in range). The time period for the data
set is from 12:57:26 to 13:00:32. In this scenario, an aircraft
proceeds laterally across the window and then turns to align
itself with the runway in preparation for take-off.
The output of the Stick-Figure tracker is closely follows
the apparent orientation of the radar returns. The output of
the other two trackers closely overlay one another. However,
the key here is that any algorithm based on using only the centroids
and a straight-line motion model will lag behind several seconds
as an aircraft makes a turn. It follows that the Stick-Figure
tracker has the ability to predict an imminent runway incursion,
whereas the other trackers would detect the incursion only after
it had occurred.
References
- Lyon, Ervin F., "Airport Surface Traffic
Automation," The Lincoln Laboratory Journal, 4:2, 1991.
- Harrison, Michael J., "Runway Incursions
and Airport Surface Traffic Automation," unpublished manuscript,
Research and Development Service, Federal Aviation Administration,
1991.
- Schwab, Charles E., "Airport Surface
Detection Equipment," Proceedings of the IEEE, vol. 73,
pp 290-300, 1985.
- Bar-Shalom, Yaakov and Fortmann, Thomas
E., Tracking and Data Association, Academic Press, 1988.
- Butler, C.A. and Hulbert, D.S., "E-3
Sensor Data Fusion Algorithm Research and Demonstration,"
Proceedings of the 1991 Symposium on Command and Control Research,
June 1991.
- Jazwinski, A.H., Stochastic Processes and
Filtering Theory, Academic Press, 1970.
The development of the Stick-Figure
tracker was sponsored in part by the Airport Surface Traffic Automation
(ASTA) program of the Federal Aviation Administration, under the
management of Mr. John Heurtley, Code ARD-50. The
algorithm is an original and proprietary approach developed by
Daniel H. Wagner Associates, Inc. and was demonstrated
on simulated data prior to this study. The actual airport runway
radar data was provided by John Heurtley. Mr. Richard Bush
of MIT Lincoln Laboratory provided assistance in interpreting
the radar data.
|
