Transcript Search Performance

UNCLASSIFIED
1st Annual Israel Multinational BMD Conference & Exhibition
SYSTEM LEVEL RADAR
SEARCH OPTIMIZATION
Presented By:
Ms. A. Gur
WALES, Ltd., Israel
System Level Radar Search Optimization -
WALES, Ltd.
UNCLASSIFIED
MAY 2010
1/20
UNCLASSIFIED
RADAR SEARCH PLANNING
PRESENTATION TOPICS
• Introduction
• The Search Patterning Problem
• BeamCover Tool Description
• Search Pattern Trade-Offs
• Summary
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
2/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
INTRODUCTION
• BMD systems rely on radars to search the sky and detect
incoming TBMs
• Search performance is key to the effectiveness of the
defense system
• The definition of the search mission offers interesting
tradeoffs between:
– Probability of detection (Pd)
– Sensor load
– Interception battlespace
• In order to probe these issues WALES, Ltd. developed the
BeamCover search patterning tool
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
3/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
BEAMCOVER OVERVIEW
• At the heart, BeamCover is a search pattern solver.
Given a search mission it attempts to find a satisfying
search pattern
• BeamCover search patterns are distributions of sensor
resources on a single beam resolution
• On this engine sits an optimization driver that pushes
the search mission to various limits according to
optimization requirements
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
4/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
THE SEARCH PATTERNING PROBLEM
– A sample of possible threat trajectories
(geometry, RCS)
– Timeline constraints
– Probability of detection (Pd) requirements
– A set of possible radar search beams
(direction, waveform)
– Resources allocated for search
• Find a beam-wise distribution of search
resources that provides the required Pd
and timeline constraints for all threat
trajectories
• The challenge here is the large number of
trajectories needed to sample the threat
(~105 for a large enemy country) and the
large number of possible beams (~104)
Elevation
• Given a search mission defined by:
Azimuth
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
5/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
BASIC TOOL PROPERTIES – RADAR, THREAT AND TIMELINE
• Geographic location, orientation, radar horizon, etc.
• A palette of dwells (waveforms) defined by their power,
duration, instrumented range, Pfa, # of range and Doppler
gates, and pulse structure
• Azimuth X Elevation grid of possible beam directions
• Beam amplitude – Gaussian with off boresight effects
• Detection process accounts for:
– Threat RCS - mean and distribution type
– Trajectory path inside beam
– Verification process
• BeamCover supports a variety of timeline constraints:
– Fixed: Detection no later than t seconds prior to threat
descending to altitude h
– Interceptor time of flight dependent: detection should enable
interception at altitude h by at least one firing unit
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
6/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
BASIC TOOL PROPERTIES – SEARCH PATTERN SOLVER
• We start with a guess search pattern (e.g. random distribution of
resources)
• Define an objective function that reflects the distance between the
required performance and the performance of the current search
pattern
• A Simulated Annealing solver is used to gradually modify the
resource distribution in an attempt to meet the search mission
requirements:
– Small quantities of radar resources are stochastically moved around
until a suitable solution is found
– The algorithm uses a combination of “good” (objective function
reducing) and “bad” (objective function increasing) moves in a way that
converges to a solution while allowing escapes from local minima
• Solver is capable of dealing with large problems (hundreds of
thousands of trajectories, tens of thousands of beams) within
several days, on a desktop PC
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
7/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
OPTIMIZATION MODES – RESOURCE REDUCTION
• The most straightforward optimization goal is to find a search
pattern that satisfies the search mission using the least amount
of resources
• We do this by reducing the available resources every time the
solver succeeds in finding a solution
Initial search mission:
• Trajectories
• Timeline constraints
• Required Pd
• Initial guess search resources
Find search pattern
that satisfies initial
mission
Reduce search
resources by a
small amount
Failed
Simulated
Annealing
Solver
Optimization
Driver
Succeed
Succeed
Legend
Adapt search
pattern to restore
Pd
Failed
Increase
search
resources
Save last
valid solution
and stop
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
9/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
MINIMAL RESOURCES EXAMPLE
• Initial mission: provide minimal engagement interval with
minimal resources
• Result: optimal search pattern, minimal resources required
=12.5%
Revisit rate
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
10/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
OPTIMIZATION MODES – INTERCEPTION WINDOW EXPANSION
• Given a fixed amount of resources, find a search pattern that
maximizes the interception windows of the different trajectories
• We do this by increasing a trajectory’s interception window by a
small amount every time the solver succeeds in finding a solution
Initial search mission:
• Trajectories
• Timeline constraints for minimal interception window
• Required Pd
• Search resources
Find search pattern
that satisfies initial
mission
Failed
Succeed
Succeed
• Choose a trajectory whose
interception window can be expanded
(not limited by horizon, discrimination
or interceptor constraints)
• Expand trajectory interception window
by a small amount
Legend
Simulated
Annealing
Solver
Optimization
Driver
Adapt search
pattern to restore
Pd
Failed
Save last
valid solution
and stop
Not enough
resources
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
11/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
AVAILABLE INTERCEPTION WINDOW EXAMPLE
• Allow search resources of 55%
Portion of trajectories
• Provide available interception window for a given defense
deployment
12.5%
Resources
55%
Resources
Interception Altitude
Revisit rate
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
13/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
ADDITIONAL CAPABILITIES –
PERFORMANCE IN A COMPLEX ENVIRONMENT
• BeamCover can compute multi sensor search patterns
• Radars are straightforwardly added by including their
beam set in the search patterning problem
• For each part of the threat we can specify the sensor or
sensors that should detect it. We thus have the flexibility
of designing a mixture of joint and individual detection
requirements
• Resource constraints can be put on the overall array and
on individual sensors
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
14/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
ADDITIONAL CAPABILITIES –
MONTE CARLO DETECTION SIMULATION
• BeamCover uses approximations to speed up pattern
computation
• To verify that the approximations are sound, an independent
Monte Carlo detection simulation was built into the tool
• The simulation is time driven:
– Execute the search pattern with an individual dwell time resolution
(10-100 msec)
– Propagate the threat position with the pattern execution time
– Draw random RCS values according to the threat RCS distribution
– Declare detection if the return signal is strong enough (including
verification)
– Declare a leaker if the threat passes its last detection time
undetected
• Repeating this process many times results in trajectories
detection statistics which can be compared to the approximate
trajectories Pd’s
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
15/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
ADDITIONAL CAPABILITIES –
MONTE CARLO DETECTION SIMULATION (Cont.)
• Monte Carlo detection statistics of a very large set of
trajectories from the threat set
Pd
• Trajectories are ordered according to their BeamCover reported
Pd
• BeamCover Pd approximation
• 98% boundaries of blue
curve’s binomial distribution
(5000 attempts)
• Monte Carlo statistics
Excellent agreement between
BeamCover’s approximation and
Monte Carlo statistics
Trajectory index
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
16/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
TOOL APPLICATIONS
• Optimal performance:
– Comparison between sensor options
– Deployment recommendations
– Search pattern design
• Impact of mission trade-offs
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
17/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
MISSION TRADEOFFS
• The search radar provides a given engagement battlespace
for a defined level of resources
• In cases of load or sensor degradation it may be necessary
to reduce performance of radar search
• There are various approaches to reallocate the sensor
search resources that will have different impact on the
defense mission
• Potential trade-offs include:
– Timeline compromise – delay detection by t
– Pd compromise
– Reduced threat set (launch sites or impact sites)
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
18/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
MISSION TRADEOFFS (Cont.)
Pd Compromise
Radar Resources
Radar Resources
Timeline Compromise
0.88
0.9
0.92
0.94
0.96
0.98
1
Probability of Detection
Engagement Timeline
In this example, Pd degradation seems more effective
than timeline degradation to reduce resources
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
19/20
SEPTEMBER 2009
UNCLASSIFIED
RADAR SEARCH PLANNING
SUMMARY
• Appropriate allocation of radar resources for search can
have a critical impact on defense system performance
• BeamCover is a flexible search patterning tool, designed to
allow the architecture planner to make the most out of the
(limited) search resources at his / her disposal
• The BeamCover tool can be used to assess the search
capability of a complex multi-radar defense architecture
• The tool is being used both to assess how to expand
Defense Architectures as well as to find algorithms for
operation under degraded performance
BMD Conf. – Radar Search Planning -
WALES, Ltd.
UNCLASSIFIED
20/20
SEPTEMBER 2009