You are at: Wagner
Home > Technologies > Data Fusion Tracking
> AWACS Level 2 Fusion
Level 2 Data Fusion for E-3
The E-3 AWACS SENTRY aircraft is the "brain" of the
modern air war for the U.S. and its allies. The aircraft has powerful
active and passive sensors and an array of Level 1 tracking algorithms
for managing the real time kinematic "picture" of the
air battle. The Air Force has an ambitious program of sensor enhancements
and software to improve this Level 1 processing and Daniel H.
Wagner Associates, Inc. is a key player in the effort to improve
Level 1 system as a participant in the Multi-Sensor
Integration arena. Our MSI algorithm was the first to be demonstrated
in ESC's Fusion Evaluation Testbed and we have consistently shown
outstanding test performance. In ongoing research for the Air
Force Research Laboratory, Wagner Associates is investigating
and implementing a number of Level 2 Data Fusion algorithms for
use onboard the E-3 AWACS.
Tools for Level 2 Data Fusion fall into two categories: Situation
Assessment (SA) and Sensor Management (SM). The functions of Situation
Assessment are twofold: (1) Operator Awareness and (2) Operator
Workload Reduction. The functions of Sensor Management are: (1)
Recommending sensor settings for sectors and subsectors based
on situation and (2) Managing complex settings for the operator.
Below is a list of topics being investigated.
In the Air Tasking Order (ATO), fighters are assigned to specific
tankers for refueling. To avoid confusion in operations with large
numbers of aircraft, the assignments and appropriate vectors could
be displayed on the operator's screen, allowing the operator to
vector an aircraft to the correct tanker quickly and reliably.
The system could also automatically detect when a fighter is approaching
the wrong tanker and alert the operator.
Air corridors provide a strong correlation for identifying
non-hostile tracks, using altitude, course, speed, and conformance
to the corridor route. A situation assessment algorithm would
compare an unidentified target's behavior to the stored corridors
and cue the MSI to assign non-hostile designations. Such cueing
should ordinarily be used in conjunction with appropriate IFF
Sign Recognition from Voice Channels
Speech recognition is now an off-the-shelf capability well
within the capacity of desktop and tactical computers. In AWACS,
operator radio commands to tracks under control could be parsed
by computer. A simple measure of (1) recognizing the call sign
from operator speech and (2) indicating the current location of
the track being called, could prevent operator mistakes caused
by incorrect track identification or simple call sign confusion.
When an unfriendly track seeks to avoid detection, it can remain
in the shadow of terrain, either as a defensive or offensive maneuver.
In a defensive maneuver, it can loiter for long periods, remaining
in radar shadow. In an offensive maneuver, it can select an attack
route that minimizes detection by virtue of terrain masking.
A template that incorporated terrain masking algorithms would
be able to match the loss of radar contact with the presence of
terrain. The use of nonlinear tracking techniques could hold a
track on the target for long periods of time without radar contact,
as long as the terrain masking persisted. A useful model for nonlinear
tracking is Monte Carlo, where large numbers of sample paths are
replicated, drawn at random from the possible paths by which the
track could traverse the terrain region.
Strike controllers often need to make instant decisions to
commit resources to a target. Depending on the type of target,
only resources with certain ordnance may be feasible. Also, a
resource must either have available fuel or must be refueled in
order to complete the mission. Larger, more complex decision aids
can be used by other command elements to pre-plan missions for
the ATO. However, when the AWACS controller has a need for an
immediate-reaction resource, a quick-information decision aid
will be sufficient to assist in identifying the correct resources.
In a quick-response mode, the operator would select a target
and resource (or use the current selection generated in the commitment
decision aid). The routing algorithm would compute the fastest
route to the target that had a computed risk below some maximum
value. The decision aid would temporarily display the route in
the operator window. The operator could change the route by moving
an endpoint or by creating and moving an existing endpoint in
the middle of one of the legs.
In certain tactical situations, multiple tracks that have been
positively identified come into close proximity and, because of
range resolution and radar sweep time, can no longer be distinguished
from one another. If one or more of these tracks are friendly,
then Mode 4 IFF interrogation can be used to re-establish positive
ID. However, Mode 4 IFF is a resource to be used sparingly, according
to standard doctrine. If a Level 2 algorithm can routinely examine
the uncertainties in track ID from the tracker/correlator output,
it could automatically order Mode 4 interrogations over limited
sectors, based on the predicted azimuth of the track(s) in question.
This would maximize the ID quality of the track picture while
minimizing the use of the scarce Mode 4 resource.
If a track's velocity is tangential to the line of sight from
the E-3, then its absolute range rate is zero, which matches the
range rate of the ground. In pulse Doppler mode, the radar will
lose such a track in what is called the "clutter notch."
If the sensor management algorithm is notified of loss of radar
contact, it can generate a stochastic motion model that represents
the possible path of the target. If it continues not to be reported,
the algorithm could switch the radar to LVD (Low Velocity Detection)
or even BTH (non-Doppler) mode for a short period of time in order
to re-acquire the critical target in its predicted sector. This
could be repeated for as long as necessary to maintain continuous
Radar automatic detection processes extract "blobs"
of energy from the radar signal return using thresholding techniques
designed to find a balance between suppressing clutter and preserving
actual target signals. Thresholds that provide adequate probability
of detection for small radar cross-sections targets in the entire
surveillance region would produce unacceptably high numbers of
false targets. Subsectors allow the AWACS operator to modify certain
radar parameters over a limited space in order to improve detection
for certain types of targets. The Level 2 Fusion Algorithm will
be able to create subsectors and nominate settings for certain
situations based on templates. The operator could also use the
same facility and create special subsectors as needed.