U.S. patent application number 12/897841 was filed with the patent office on 2012-06-21 for genetic algorithm enhancement of radar system survivability.
This patent application is currently assigned to U.S. Government as Represented by the Secretary of the Army. Invention is credited to PHILIPP ARTHUR DJANG, EDWARD FRIDAY, GENE E. HERRIMAN, FRANK LOPEZ, EDUARDO RUILOBA, JR..
Application Number | 20120154197 12/897841 |
Document ID | / |
Family ID | 46233681 |
Filed Date | 2012-06-21 |
United States Patent
Application |
20120154197 |
Kind Code |
A1 |
DJANG; PHILIPP ARTHUR ; et
al. |
June 21, 2012 |
GENETIC ALGORITHM ENHANCEMENT OF RADAR SYSTEM SURVIVABILITY
Abstract
A process for enhancing radar system survivability against a
threat includes the construction of multiple computational radar
function chromosomes with each of the chromosomes having values for
geometric configuration and operational parameters for at least one
radar station emitter and at least one decoy. A genetic algorithm
analysis is performed on a computer including a crossover operator
and a mutation operator to determine a fitness value for multiple
chromosomes against the threat. By configuring and operating the
radar system based on a survivability value for the fitness value
of the genetic algorithm analysis, radar system survivability
against a threat is enhanced. A radar system employing genetic
algorithm analysis is also provided.
Inventors: |
DJANG; PHILIPP ARTHUR; (Las
Cruces, NM) ; LOPEZ; FRANK; (El Paso, TX) ;
HERRIMAN; GENE E.; (El Paso, NM) ; RUILOBA, JR.;
EDUARDO; (El Paso, TX) ; FRIDAY; EDWARD; (Las
Cruces, NM) |
Assignee: |
U.S. Government as Represented by
the Secretary of the Army
Adelphi
MD
|
Family ID: |
46233681 |
Appl. No.: |
12/897841 |
Filed: |
October 5, 2010 |
Current U.S.
Class: |
342/13 |
Current CPC
Class: |
G01S 13/87 20130101;
G01S 7/36 20130101 |
Class at
Publication: |
342/13 |
International
Class: |
G01S 7/42 20060101
G01S007/42 |
Goverment Interests
GOVERNMENT INTEREST
[0001] The invention described herein may be manufactured, used,
sold, imported or licensed by or for the United States Government.
Claims
1. A process for enhancing radar system survivability against an
anti-radiation missile threat comprising: constructing a plurality
of computational radar function chromosomes, each of said plurality
of chromosomes comprising N gene values for geometric configuration
and operational parameters for at least one radar emitter station
and at least one decoy; performing a genetic algorithm analysis on
a computer using at least a crossover operator and a mutation
operator to determine a fitness value for each of said plurality of
chromosomes against the threat; and configuring and operating the
radar system based on a protective survivability value against the
threat for said fitness value of the genetic algorithm
analysis.
2. The process of claim 1 wherein each of said plurality of
chromosomes includes gene values for parameters selected from the
group consisting of: radar emitting element failure percent, decoy
number, decoy radiating power, decoy blink, and decoy activation
time.
3. The process of claim 1 wherein each of said plurality of
chromosomes has a non-zero distance between said at least one decoy
and said at least one radar emitter station.
4. The process of claim 1 wherein said crossover operator operates
only between two or more of said plurality of chromosomes having
equivalent decoy numbers.
5. The process of claim 1 wherein said crossover operator is a
homogeneous multiple crossover operator having a crossover
frequency of between an average 2 and N-1 genes.
6. The process of claim 1 wherein the fitness value is obtained by
scoring each of said chromosomes; proportionately applying at least
said crossover operator and said mutation operator with a given
frequency to said plurality of chromosomes; and generating
therefrom a progeny generation of chromosomes wherein an elitist
best performing chromosome is maintained in said progeny
chromosomes.
7. The process of claim 1 further comprising performing a second
genetic algorithm analysis against a second threat to yield a
second threat protective survivability value against the second
threat for the radar system; and storing said protective
survivability value and said second protective survivability value
in a computer storage library with recall and implementation of one
of said protective survivability value and said second protective
survivability value in response to information input about a
developing real world threat.
8. The process of claim 1 further comprising providing a
communication receiver to said computer to provide input
information about a developing threat to facilitate searching said
computer storage library.
9. The process of claim 1 wherein said genetic algorithm analysis
is stochastic.
10. The process of claim 1 wherein said mutation operator replaces
one of said gene values with a mutated value chosen from a feasible
range of values for said gene value.
11. The process of claim 1 wherein an operator application
frequency for said crossover operator and said mutation operator
changes during the determination of said fitness value for each of
said plurality of chromosomes.
12. The process of claim 11 wherein said operator application
frequency for application of said crossover operator decreases and
said mutation operator increases as a rate of change of said
fitness value decreases between successive chromosome generations
during said genetic algorithm analysis.
13. The process of claim 1 wherein said genetic algorithm analysis
allocates additional application of said crossover operator and
said mutation operator to one of said plurality of chromosomes
having an above average score of said fitness value.
14. The process of claim 1 further comprising storing a limited
subset of said plurality of chromosomes for use in subsequent
chromosome generations of said genetic algorithm analysis.
15. The process of claim 1 wherein said fitness value is determined
by a weighted summation of factors including at least two of:
number of radar hits, number of decoy hits, and near miss distance
to one of said at least one radar emitter station and said
decoy.
16. The process of claim 15 wherein the near miss distance value
associated with said protective survivability value for said
fitness value is at least 800 meters.
17. A radar defense system comprising: a radar emitter station; a
decoy placed a non-zero distance from said radar emitter station; a
communication link between said radar emitter station and said
decoy; a computer coupled to said radar emitter station and
performing a genetic algorithm analysis to determine a fitness
value for a survivability operational parameter chromosome for the
system against an anti-radiation missile threat, said computer
having a computer storage storing a plurality of system operational
parameter chromosomes determined under different threats and the
fitness value for a survivability operational parameter chromosome;
and a communication receiver receiving developing threat data and
providing the data to said computer to facilitate selection of an
optimal one of said plurality of fitness values for implementation
by said radar emitter station and said decoy.
18. The system of claim 17 wherein said optimal one of said
plurality of fitness values controls parameters inclusive of
radiating power of said decoy, blink of said decoy, activation time
of said decoy, and activation time profile of said radar emitter
station.
19. The system of claim 17 wherein said genetic algorithm analysis
employs a parent roulette wheel methodology with an elitist
retention scoring to determine said survivability operational
parameter chromosome.
20. The system of claim 17 wherein said fitness value is determined
by a weighted summation of factors including at least two of:
number of radar hits, number of decoy hits, and near miss distance
to one of said at least one radar emitter station and said decoy.
Description
FIELD OF THE INVENTION
[0002] The present invention relates to improving radar survival
against a threat, and in particular to the use of a genetic
algorithm to rapidly converge on radar configurations and
operational parameters to improve survivability.
BACKGROUND OF THE INVENTION
[0003] Air and missile defense radar systems provide early warning
as to both battlefield and theater threats. A well designed air and
missile defense radar system provides sufficient advance warning
for ground personnel to take evasive actions, interceptor aircraft
or missile assets can be vectored towards the threat, or electronic
jamming devices employed. Owing to the effectiveness of air and
missile defense radar systems against an aggressor, conventional
air attack doctrine includes an anti-radiation missile (ARM) attack
component to blind a defender as to an airborne threat and create
air and/or missile attack corridors. Even in instances where a
defender retains air superiority, the threat of ARM attack persists
owing to growing prevalence of mobile ground fired ARM and drones
deploying ARMs.
[0004] A problem with defending an air and missile radar defense
system against a potential threat is the complexity of the problem.
To model a radar defense system and its perceived performance
against an ARM threat involves approximately 200 interdependent
variables resulting in a stochastic computation that is so complex
that to analyze all the valid combinations involves millions of
computational runs. As a result, even with high speed computational
resources, radar system operational optimization in the face of an
evolving threat is currently so slow as to effectively be
intractable. Genetic algorithms have been shown effective in
optimizing peak to side lobe ratio radar ambiguity functions. T.
Bucciarelli et al., Proceedings of the 8.sup.th Intl. Conf. on
Signal Processing Applns and Tech., Miller Freeman, Vol. 2, pages
1862-1866 (1997). Genetic analysis has also been used in the
identification and design of optimal teams of sensors to detect
enemy radars using genetic analysis. Yilmaz et al., "Evolving
Sensor Suites for Enemy Radar Detection". However, these past
efforts have not addressed issues related to improving radar
defense system survivability against ARMs.
[0005] Currently, radar emitting beacon decoys are used to enhance
radar survivability by presenting radar signal clutter to deceive
an ARM. Typically, a decoy emits operational radar-like signals, so
as to distract or confuse an ARM. However, the geometric
configuration of a decoy or group of decoys in a radar system is
complex. Likewise, the operational conditions of a radar and the
related decoys to blunt a given ARM threat is a complex problem
that may evolve in real time based on the threat scenario.
[0006] Thus, there exists a need to enhance configurational and
operational parameters of a defense radar system against an ARM
threat. There further exists a need to calculate configurational
and operational parameters in an efficient way with resort to
genetic analysis to maximize a fitness function for a given set of
configurational and operational parameters using a survival of the
fittest approach. Genetic algorithms are not known to have
previously been applied to the problem of enhancing radar system
survival against ARM threat.
SUMMARY OF THE INVENTION
[0007] A process for enhancing radar system survivability against a
threat includes the construction of multiple computational radar
function chromosomes with each of the chromosomes having values for
geometric configuration and operational parameters for at least one
radar station emitter and at least one decoy. A genetic algorithm
analysis is performed on a computer including a crossover operator
and a mutation operator to determine a fitness value for multiple
chromosomes against the threat. By configuring and operating the
radar system based on a survivability value for the fitness value
of the genetic algorithm analysis, radar system survivability
against a threat is enhanced. A radar system employing genetic
algorithm analysis is also provided.
BRIEF DESCRIPTION OF THE DRAWINGS
[0008] FIG. 1 is a schematic of a radar system in an ARM launch
environment;
[0009] FIG. 2 is a block diagram approach to an exemplary
simulation;
[0010] FIG. 3 is a schematic of an exemplary run-scoring
methodology; and
[0011] FIG. 4 is a bar graph of a genetic algorithm ARM simulation
results obtained for successive chromosome generations to show
fitness value evolution.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0012] The present invention has utility in providing a process for
enhancing radar systems against an anti-radiation missile (ARM)
threat. A novel genetic algorithm is used with a simulation of
survivability for a given chromosome, to obtain configurational and
operational parameters with enhanced survivability. Previous
approaches have used decoys that mimic radar, where survivability
of the radar then depended on the location and mimicked behavior of
each decoy. However, the complexity of these systems is so large
that a full analysis of the problem has previously been
intractable.
[0013] An inventive hybrid genetic algorithm is provided that
allows optimization of the operational states, locations, and
number of radars and decoys used to enhance survivability of a
radar station to an evolving threat.
[0014] In a representative example, a novel hybrid genetic
algorithm is designed to allow optimization of the survivability of
at least one radar station emitter and at least one decoy against
an anti-radiation missile (ARM) attack. The algorithm provides
configurational and operational parameters of the radar and
decoy(s). Optionally, the number of decoys and decoy configuration
relative to the actual radar station emitter is provided for design
of a new station. Alternatively, a decoy number, decoy
configurational position, operational functions, or combinations
thereof are fixed to provide an enhanced survival chromosome for a
deployed radar system. This ability to rapidly compute a chromosome
of radar system parameters allows for evolution in system operation
in a timeframe sufficient to counter a new threat as it develops or
storage in a digital library of survivability values for a given
threat. The configurations and operational parameters are
determined by a novel genetic algorithm, and a specialized scoring
system provides feedback to the genetic algorithm as to the fitness
value performance of any given chromosome of radar system
configuration and operational parameters to propagate a new
generation of chromosomes with operator usage to improve chromosome
fitness. By way of example, the flight, fuse point of an ARM, and
survivability characteristics of a radar station emitter and
decoy(s) are stochastically simulated to score the fitness of a
tested chromosomal solution.
[0015] The results of the genetic analysis are used to propagate
genes having fitness in subsequent generations of solution
chromosomes to enhance the survivability of a radar and decoy(s)
against the threat. By scoring chromosomes against different
evolutionary pressures, different survival enhancing chromosomes
are obtained against various threats illustratively including
survivability against various threat scenarios such as a ripple
fired multiple missile attack, survivability against a missile
attack as a function of missile launch parameters, and
survivability against a plurality of missiles fired from various
locations, and commando launched ARM attack proximal to the radar
system. Thus, when a threat evolves, a chromosome adapted to have a
protective fitness value is implemented from a library of
chromosomes, each optimized under different evolutionary threat
pressures.
[0016] As used herein, a chromosome is defined as a function
c.sub.i(x) where i equals 1, . . . N where N is the dimension of
the population of possible solutions; x is the configurational and
operational parameters of the radar system and illustratively
includes values for radar emitting element failure percent, decoy
number, decoy radiating power, decoy blink, and decoy activation
time. Each of the configurational and operational parameter values
in a chromosome is described herein synonymously as a gene.
[0017] According to the present invention, an initial population of
chromosomes having certain genes, each individually fixed or
bounded by a given range based on real world operation is scored
individually to provide a fitness value against a user input
threat. The bounding of values is critical to obtaining useful
fitness values for radar system operation. As a counter example,
optimal survivability against ARM threat converges to the system
being inactivated permanently. While a radar station with no
operation is a valid theoretical solution, in practice allowing a
chromosome to converge to this solution without a radar active
range is of no practical value. Chromosomes having a fitness that
enhances survivability against the threat better than the fitness
of other chromosomes, are propagated preferentially into the next
generation of chromosomes with the process repeated to iteratively
evolve chromosomes with survivability against a given threat. To
facilitate rapid evolution of a fitness value for a radar station
configuration operational parameter chromosome, genetic operators
are applied during the course of the evolutionary genetic algorithm
analysis to facilitate offspring inheritance of genes enhancing
overall fitness. Genetic operators used in an inventive genetic
algorithm analysis include a crossover operator C that binds
chromosomes to generate new individuals. An example of a single
homogeneous crossover operator function for a two chromosome case
is provided in Formula I that leaves the progeny chromosomes of
equal length as the parent chromosomes.
c 1 : abc - def C c 2 : ghi - jkl [ abc def ghi jkl ] -> .times.
c 1 ' : abc - jkl c 2 ' : ghi - def ( I ) ##EQU00001##
where c.sub.1 and c.sub.2 are parent chromosomes 1 and 2 and
progeny chromosomes after operation of the single homogeneous
crossover operator C are c.sub.1' and c.sub.2'. It is appreciated
that while biological genetics only allows for the operation of a
given operator between two chromosomes, the computational genetics
of the present invention are not so limited and as such an
inventive genetic algorithm operator such as the crossover operator
described above as well as other operators described herein are
readily applied in ways non-analogous to biology to evolve a
progeny generation of fitness scorable chromosomes. According to
the present invention, a multiple crossing homogeneous crossover
operator is preferably applied with a frequency input by a user,
with that frequency defined as a probability of crossover. The
probability of crossover is preferably according to the present
invention a variable dependent upon the rate of change in the
fitness value as the chromosomes evolve. A homogeneous multiple
crossover operator creates more than the single crossover of
Formula I in progeny chromosomes of like length.
[0018] Another operator used to impart diversity to a chromosome
population during fitness evolution is a mutation operator, M. The
mutation operator functions to place a given gene with a different
value from within a range for the given gene or change a gene
condition, for example to exchange "on" for "off'. A stochastic
mutation operator according to the present invention is fixed to
allow real world values for a given gene. The mutation operator is
applied with a user supplied probability, P.sub.M. Preferably, the
mutation operator probability is applied as a function of the rate
of change in fitness value between succeeding generations of
iterative chromosomes.
[0019] Iterative generations incorporate parent fitness values
using the parent roulette wheel method in which the area of a
circle, corresponding to a probability of 1, the wedge area of the
circle encompassed by a particular chromosome being proportional to
the fitness value of that chromosome and as such a greater
likelihood of contribution to the next generation of chromosomes.
The parent roulette wheel method retains population size between
generations. In addition to the stochastic operator of crossover
which is a subset of recombinatorial stochastic operators, a small
probability random mutation is introduced.
[0020] Additional operators that are optionally applied to a
population to facilitate convergence to a survivability fitness
value optionally include an inversion operator that swaps gene
values in a single chromosome around an inversion plane, a clone
operator that copies a chromosome without any change and
effectively doubles the area in the roulette wheel, a zap mutation
that changes the value of a gene or part of a chromosome to another
value, and a creep mutation that changes the value of a gene by
plus or minus one unit.
[0021] Through the use of an inventive genetic algorithm tailored
to complexities of radar system survivability, an inventive genetic
algorithm provides rapid convergence to survivability fitness value
chromosomes even though the underlying search space of radar system
survivability variables is not completely understood. As a result,
an inventive genetic algorithm provides superior conversion to
other search techniques or calculus based techniques such as
Fibonacci sets and sorting; enumerative techniques such as dynamic
programming, depth first searching (DFS), and breadth first
searching (BFS); and other guided random search techniques
inclusive of simulated annealing and random walk extrema
optimization.
[0022] The fundamental basis for genetic algorithm analysis
according to the present invention is found in the aforementioned
references. A scoring system uniquely detailed herein tests
chromosome fitness with distinct criteria unique to a radar system
survivability simulation.
[0023] A novel genetic algorithm is developed to interact with the
radar simulation to optimize survivability. Each chromosome
included configuration and parameter setting genes. The simulation
is used to create a score based on the miss distance from the radar
and the decoys. The score is then used as a measure of the fitness
of the chromosome, the term "fitness value" being analogous to
survival of the fittest in evolutionary genetics. The score in turn
is used to provide an area proportional to a chromosome fitness for
use in a roulette wheel selection of chromosomes for propagation
into the next iterative generation. A fitness value is protective
when the fitness value achieves at least survival of the radar
itself and preferably components decoys against a given threat.
[0024] Special chromosomes are used including genes that are the
features and behaviors of the radar and decoys. Further, special
operators were used to ensure feasibility of the performance of the
genetic algorithm.
[0025] A partial example of an inventive chromosome is shown below
in Table 1.
TABLE-US-00001 TABLE 1 Radar Emitting Number Decoy Decoy Decoy
Element Failure of Decoys Radiating Power Blink On Time 5% 2 10 KW
True 1.3 sec
In simulations other parameters are included, but these are not
important for illustration of the novel aspects of this example.
The radar simulation is used to evaluate a score based on the
parameters included within the chromosome. The evolutionary
stressor used to evolve the genetic solution in this simulation is
a pre-determined scenario. The scenario consists of a single ARM
attack with the following initial conditions: launch angle of 190
deg off-boresight to the radar, launch height of 10ft above sea
level, launch distance of 500 miles from the radar, and with the
ARM seeker locked on to the radar at launch. All missile and launch
platform aerodynamics were appropriately modeled. The genetic
algorithm then selects the fitter chromosomes using the score
determined by the simulation. The genetic algorithm included a
novel crossover operator and a novel mutation operator. The
operators are restricted so as to not create unfeasible or illegal
combinations, within specific rules. For example each decoy must
have at least a non-zero (X,.sub.n+Y.sub.n) location so that
chromosomes specifying a number of decoys should have corresponding
locations in their associated genes, where n is the number of each
decoy in a system.
[0026] The crossover operator employed multipoint homogeneous match
crossovers. Table 2 below illustrates a representative example.
TABLE-US-00002 TABLE 2 Radar Emitting Decoy Element Radiating Decoy
Decoy Chromosome Failure X/Y Decoy 1 Power Blink On Time Parent 1
5% 2400/1600 10 KW True 1.3 sec Parent 2 0% 1400/900 5 KW False 0.0
sec After Multi-Point Crossover Child 1 0% 2400/900 5 KW True 1.3
sec Child 2 5% 1400/1600 10 KW False 0.0 sec
The two parent chromosomes are selected for crossover based on
rules such as random selection or based on a feature such as a like
number of decoys. According to the present invention, preferably
the crossover operator involves multiple crossing points with a
crossover occurring on average between an average 2 to N-1 genes of
the chromosome. By way of example, in a 200 gene radar system
configuration and operational chromosome between 2 and 199
crossovers occur across the chromosome per application of the
operator.
[0027] The mutation operator employed a restricted allele set
mutation operator for a gene. The mutation operator selects from a
finite set of values to replace a current value within the gene.
Table 3 below illustrates a representative example.
TABLE-US-00003 TABLE 3 Radar Emitting Decoy Element Radiating Decoy
Decoy Chromosome Failure X/Y Decoy 1 Power Blink On Time Parent 1
5% 2400/1600 10 KW True 1.3 sec After Mutation Child 1 3% 2400/3500
15 KW True 1.3 sec
The inventive mutation operator is in contrast to that commonly
employed in classical genetic analysis in that any radar
operational parameters limited to specific discrete values in an
inventive mutation operator are customized to operate on a specific
gene and randomly replace that gene with a new value from a
discrete set of feasible values.
[0028] Preferably, the chromosome reproduction scheme between
generations differs from the classical roulette wheel fitness
proportionate reproduction through also including an elitist
strategy that retains the best performing chromosome unchanged in
the progeny generation.
[0029] In a conventional genetic algorithm, the relative frequency
of operator application remains fixed. Typical conventional
frequencies include 75% crossover and 10% mutation. In any
generation, 75% of the chromosomes are selected for crossover and
10% are selected for mutation. According to the present invention,
the frequency operator application is preferably adapted based on
the rate of improvement in the measurement of fitness value between
chromosome generations. Preferably, if the rate of improvement
slows, the proportion of chromosomes selected for mutation is
increased. More preferably, the frequency of crossover operations
also decreases as the rate of fitness improvement slows.
[0030] In examples of the present invention, a hybrid reproduction
scheme is used. This scheme allocates additional trial through
chromosomes with higher fitnesses. The number of additional trials
may, for example, be proportionate to the measured fitness. The
best found chromosomes from each generation were retained: an
elitist strategy. This approach allows efficient exploration of a
search space, and exploits high performance information by
retaining the best chromosome. A novel objective function and
scoring mechanism was created for the hybrid genetic algorithm. The
objective function and scoring mechanism connects the missile
simulation to the genetic algorithm by providing performance
feedback to the genetic algorithm. The objective function algorithm
takes into account the missed distance between the missile and the
radar and/or decoys, while minimizing the number of decoys. The
missile simulation was a stochastic simulation, and a statistically
significant number of replications were used to evaluate the
performance of each chromosome. Control programs were created to
collect and analyze the results. These were used by the genetic
algorithm to conduct an evolutionary based search for optimized
survivability of the radar and decoys.
[0031] Hence the approach used in this example used a number of
novel features. The genes used included the configuration and
parameter value for the radar and decoys. A homogenous match
multipoint crossover operator was used, in which chromosomes with
similar or homogeneous characteristics, for example an equal number
of decoys, were selected for multipoint crossover. This is the
swapping of alleles from parent chromosomes to create children
chromosomes. A novel mutation operator, a restricted set allele
mutation operator, was used. As radar parameters were limited to
specific values, the mutation operator recognized alleles, and
randomly drew a new value from a set of feasible values.
[0032] The reproduction scheme included an elitist strategy. This
included fitness proportionate reproduction and retention of best
performing chromosomes. The combination of both methods is rarely
used.
[0033] The operator frequency use was a further novel approach. The
rate of improvement in the objective function was tracked, and this
information was used to modify the percentage of crossover and
mutation. For example, when the rate of improvement decreased, the
percentage of crossover was decreased, and correspondingly the
percentage of mutation was increased.
[0034] FIG. 1 shows a typical threat and radar system
configuration, in which a missile 20 is launched from launch point
10 and follows flight path 12 to an impact point 14. The impact is
proximate to a radar emitter station 16, a first decoy 18 and a
second decoy 18'. A computer 24 is coupled to a station 16 to
provide a computer storage library of fit radar system operational
chromosomes and/or a rapid inventive genetic algorithm analysis. A
communication receiver 26 is provided to input information about a
developing ARM threat and in library search of fitness values or
genetic analysis search. In a typical configuration a
communications link 22 between the radar emitter station 16 and
decoy 18 and a second communication link 22 between station 16 and
18' allows decoy pulses to mimic pulses 25 produced by the radar
emitter station 16. While FIG. 1 depicts the radar emitter station
16 and two decoys 18 and 18t it is appreciated that an inventive
radar system optionally includes multiple such radar emitter
stations, more than two decoys or a combination thereof. While the
inclusion of one decoy in a radar system is essential to enhancing
system survivability against an ARM threat, more than one decoy is
shown by the present invention to enhance survivability. A benefit
of an inventive genetic algorithm analysis is identifying when the
inclusion of an additional decoy provides enhanced survivability as
well as such an inclusion affording a diminishing enhancement and
survivability relative to the expense of such an additional decoy.
In FIG. 1, the displacement distance between a radar emitting
station 16 and decoy 18 is denoted by vectors X.sub.1 and Y.sub.1
while those of the second decoy 18 are denoted by X.sub.2, Y.sub.2,
and Z.sub.2. The inclusion of a vertical displacement Z.sub.2 of a
decoy 18' relative to the radar emitter station 16 is in
recognition of radar station deployment topography or the usage of
an airborne decoy. Additionally, it is appreciated that the decoy
pulses emitted by decoy 18 and second decoy 18 are readily varied
as a function of radiating power, decoy blink, and on time as well
as geometric displacement from station 16.
[0035] To illustrate an inventive process, the threat scenario of
FIG. 1 is used to construct a fitness test based on a set of input
values as to characteristics of ARM detection time, ARM operational
classification, and ARM initial guidance. Preferably, this threat
data reflects actual experimental inputs to provide an ARM
simulation. While the scenario depicted in FIG. 1 represents a
prototypical threat, it is appreciated that multiple threat
scenarios are optionally modeled in an inventive genetic algorithm
analysis to afford a library of radar system operational parameters
assuming a fixed decoy configuration relative to a radar emitter
station so that any number of possible threats as detected and
classified is used to select fit radar system operational
chromosomes. Threat scenarios beyond that depicted in FIG. 1
include multiple ARM launches inclusive of possible variations in
launch distance, location, and classification.
[0036] A given ARM simulation is used as the selective pressure in
the inventive genetic analysis process based on the total number of
emitter radar station hits, the number of decoy hits by an ARM, and
nearest miss distance (NMD). For a given chromosome, the number of
decoys in operation is either fixed at a constant value or allowed
to evolve for the purpose of optimization in the building of a new
radar station. A genetic algorithm running score methodology
according to the present invention to score the fitness value of a
given chromosome was a summation of radar hits, decoy hits, near
miss distance and optionally the number of decoys used, with each
value preferably weighted by a coefficient. Exemplary of these
scoring coefficients are 1,000 times the number of emitter radar
station hits, 100 times the number of decoy hits, 10 times the
number of decoys used, and a unitary coefficient for near miss
distance. It is appreciated that the selection of weighting
coefficients directly affects the selection pressures on a given
set of configurational and operational parameters for a radar
system. The scoring scheme allows modification based on known
priorities, such as relative value and/or cost of radars and
decoys. Blast effect simulation includes assumptions and preferably
experimental or intelligence data as to fragmentation damage, over
pressure damage and other results associated with an ARM impact so
as to provide realistic selection pressures on the evolution of
configurational and operational chromosomes for a radar
station.
[0037] A simulation was developed to determine survivability of the
radar and decoy for a given impact point. Raw data was generated by
high fidelity radar simulations, using actual data when available.
Data is converted into time packets allowing pulse-to-pulse
characteristics such as time, width, frequency, and beam position
to be simulated. Various files are generated for specific
operational scenarios. The data allowed creation of synchronous
decoy pulses.
[0038] FIG. 2 represents a possible approach to simulation. Box 30
corresponds to ARM detection, ARM classification, and initial ARM
guidance. Box 32 corresponds to inclusion of flight dynamics, and
radar and decoy countermeasure (CM) effects. Box 34 corresponds to
simulation of ARM trajectories using a computer. Box 36 corresponds
to simulation of the blast effect, as a function of impact point
relative to radar and decoy locations. The simulation includes the
effects of fragmentation damage and over pressure damage.
[0039] A full simulation can include more than 200 interdependent
variables. However even a subset of genes shown in the following
tables provides enhanced system survivability while identifying low
import variables that are readily dropped from the analysis to
accrue computational speed of fitness value evolution. Hence, other
non-inventive optimization approaches, configurations of radars and
decoys are not analytically determined, but are positioned using
expert opinion. However this approach is inefficient, and the
inventive process efficiently finds an optimum radar system
configuration based on a single threat scenario or a weighted
likelihood of possible threats.
[0040] FIG. 3 illustrates the run scoring methodology used with
reference to the threat depicted in FIG. 1. The radar 16 is located
at an origin, a decoy 18 is located at coordinates X.sub.1,
Y.sub.1, and the fuse point of the missile is located at point XY.
The nearest miss distance (NMD) is the radial distance from the
incoming missile's fuse point to the nearest emitter (decoy or
radar). In this example, NMD is set to a maximum value of 9 with
the maximum rewarded distance (greater than or equal to 2000
meters). NMD is set to a minimum value of 0 with the minimum
rewarded distance (0) and intermediate distances are scaled to a
value between 0 and 9.
[0041] By creating distinctly separated tiered values for the
subgroups, the subgroups may be tiered by level of importance
and/or cost. In the latter example, the radars are considered far
more valuable than that of the decoys. The relative values given to
radar hits and decoy hits can be adjusted according to the relative
cost and/or value accorded each and adjusted as relative values are
changed by cost, durability, ARM development and other illustrative
external data.
[0042] A preliminary analysis is performed in which radar emitter
features and decoy features are genetic at specific values in each
starting chromosome. The radar ARM simulation used is preferably
stochastic, and multiple accorded simulations are used to estimate
the fitness of each chromosome. A statistical analysis of over
70,000 ARM simulations indicated that decoys are essential to radar
survivability. Without a decoy, radar survivability is 5.1% per
FIG. 1 threat. In contrast, with 3 decoys survivability increased
to 98% per FIG. 1. One decoy led to survivability of 47% per FIG. 1
threat. However, the combinatorial nature of the problem precludes
a complete analysis of the parameter space, and thus the
improvement yielded by the inventive genetic algorithm.
[0043] FIG. 4 is a bar graph of genetic algorithm results showing a
change in fitness score as a function of generations exposed to
threat evolutionary pressure. The genetic algorithm driven missile
simulation produced configurations that generated large ARM miss
distances. Decoy deployment characteristics were found that
resulted in radar and decoy survivability. The evolutionary driven
process identified high quality solutions more efficiently than
simply calculating parameters within the entire search space by
other calculus based, enumerative, or other guided random search
techniques. In FIG. 4, the following terms are used based to test
various system operation parameters. Low ARM Signature Surveillance
(LASS) is a waveform pattern of operation that a radar optionally
employs. Notches refer to single features of low field strength
within the radar pattern, generated by a special algorithm within
the radar emission control logic. "Async" refers to a flag variable
which indicates that the decoy's pulses are asynchronous to the
radar pulses. Decoy radiated pulse or "DRP" is a variable that
holds the value in Watts of a decoy's radiated power. Blink Mode
refers to a radiation mode in which intermittent radiation is
emitted by decoys. Synchronous pulse radiation intensity or "SPRI"
refers to a radiation emitting mode held constant during a decoy's
radiative time in seconds when turned ON. Asynchronous delay time
or "ADT" refers to a radiation mode in which intermittent radiation
is emitted which holds the decoy's pulse constant with a delay
times in seconds.
[0044] Hence, embodiments of the present invention include a highly
customized genetic algorithm to optimize radar survivability.
Associated software code was developed for the genetic algorithm
and is believed to be the first time a genetic algorithm is linked
to an anti-radiation missile simulation.
[0045] Patent documents and publications mentioned in the
specification are indicative of the levels of those skilled in the
art to which the invention pertains. These documents and
publications are incorporated herein by reference to the same
extent as if each individual document or publication was
specifically and individually incorporated herein by reference.
[0046] The foregoing description is illustrative of particular
embodiments of the invention, but is not meant to be a limitation
upon the practice thereof The following claims, including all
equivalents thereof, are intended to define the scope of the
invention.
* * * * *