Space-Based Systems for Missile Surveillance

Space-Based Systems for Missile Surveillance

D. G. Lawrie and T. S. Lomheim

Aerospace uses sophisticated analysis and simulation tools to design systems, assess their performance, and link design and performance to system cost.

Early warning against missile attack is a key mission for military planners dealing with missile defense systems, the subject of considerable international debate. A space-based infrared surveillance system can provide such early warning. Enhancing its timeliness and usefulness to guarantee high system performance depends on accurate appraisal of the characteristics of the background against which the target will be detected and how these characteristics influence sensor design and performance. Optimum sensors must be deployed to ensure high system performance.

The cost of optimum system performance, however, is being carefully scrutinized by decision-makers as part of the ongoing process of Department of Defense (DOD) acquisition reform that has distinguished the procurements of the past decade. Indeed, system cost has become a critical factor in decisions regarding DOD acquisitions. The inclusion of system costs in architecture studies represents a significant and important extension of the traditional role of The Aerospace Corporation in support of the Air Force Space and Missile Systems Center.

Scan pattern of sensor for theater missile surveillance; the scanner field of regard typically covers most of the Eurasian landmass. The instantaneous field of view of the sensor is shown as FOV_focal plane, which depends on both the detector height and the array length perpendicular to the scan direction. The FOV_focal plane is scanned a length FOV_in-scan over three successive adjacent "scan patches" to cover a cross-scan extent, FOV_cross-scan.

The costs associated with the deployment of a surveillance satellite are closely related to the payload mass, both in regard to the payload itself and that of the launch vehicle required to lift it and its satellite platform into, for example, a geostationary orbit. The best performing sensor has the highest resolution and the best background clutter suppression, but is the heaviest and therefore most expensive.

The study presented in this article deals with a constellation of space-based infrared surveillance sensors as an example to show the link between system performance and system cost. Simulation tools are used to facilitate the trades required for optimizing sensor designs to meet mission requirements and to indicate the best approach for minimizing system costs. Sensor resolution (the ability to see detail) is varied over three fixed-parameter designs in order to examine the impact of resolution on background clutter and, hence, system performance. The most valuable aspect of this example is the quantitative link between system-level performance and payload mass over conditions of variable clutter.

Aerospace Simulation and Modeling Tools

Aerospace regularly provides quick-response assessments of a variety of space-based-sensor concepts. These assessments, underpinned by reasonably detailed sensor design constructs, accurately determine system performance. The constructs are important for estimating sensor (and ultimately space-segment) mass, power, volume, and cost, and for evaluating associated technology risks for sensor subsystems and their components.

A variety of analysis and simulation tools are used to assess sensor performance. An analytical approach is usually adequate, unless the background-scene structure interacts with the sensor to create a significant component of the system noise (i.e., clutter). In these cases, the approximations required for an analytic approach are often violated; for example, cloud edges and land-and-sea interfaces frequently distort the normal background amplitude distribution. When this happens, detailed simulations of the spatial structure of the background scene, pixel by pixel in the focal plane, must be incorporated into the analysis for an accurate assessment of the sensor's performance. If the emphasis is on the system performance of a constellation of sensors, such a level of detail has generally been viewed as too costly and time-consuming.

Measuring the Impact of Background Spatial Structure

Aerospace has developed methodologies for quantifying the effect of the background spatial structure on the performance of space-based infrared sensors. The results are coupled with sensor-design constraints and mission performance tools to allow high-level systems-engineering trades that provide insight into relationships between cost and performance. In effect, high-fidelity sensor and phenomenology (target and background) models generate constraints and databases for use within constellation-level simulations, enhancing their accuracy. This integrated simulation capability supports sensor and system trades for a number of space-based, infrared surveillance system studies, including those dealing with theater-missile warning.

Steps, decisions, and trade-offs in the process for deriving the specifications of an infrared space-based sensor system for detecting missile launches against highly structured backgrounds. In most cases, intermediate and overall iterations are required.

An early version of this capability was used in 1994 to support the Space-Based Infrared System (SBIRS) Phenomenology Impact Study, conducted by Aerospace in collaboration with the Massachusetts Institute of Technology Lincoln Laboratory. The study recommended collecting background characteristics, a strategy that ultimately involved the Miniaturized Space Technology Initiative 3 and Midcourse Space Experiment satellite experiments, as well as background observations from a high-altitude aircraft. The phenomenology database that the study generated was later made available to the SBIRS High and Low components. SBIRS High refers to a constellation in high orbit for full Earth surveillance; SBIRS Low is a constellation in low orbit for detection and precise tracking of postboost objects, such as reentry vehicles. The simulation capability was also used during the 1994 Surveillance Summer Study as a tool for developing the system requirements for the SBIRS program.

Infrared Sensor Design

An evaluation of a space-based infrared surveillance architecture for early missile warning in two potential theaters of operation will illustrate the models and analysis procedures for assessing sensor design and performance. System performance is derived for three generic infrared scanning-sensor designs with varying degrees of spatial resolution of 1.8, 2.6, and 3.6 kilometers, with parameters specified for potential variations and uncertainties in the background structure. The sensor designs were analyzed in parallel to determine sensitivity, subsystem requirements, mass, and power. The results provide insight into the cost of the uncertainties in phenomenology, in terms of sensor mass.

sensor requirements for hypothetical surveillance system

Sensor requirements for a hypothetical theater-missile surveillance system. The flowchart illustrates the process that links the sensor requirements to steps that determine the infrared sensor design and finally the sizes of the individual infrared sensor subsystems. Designing a sensor for a single mission and a limited set of engagement geometries is straightforward. However, surveillance systems must operate against a wide variety of viewing geometries, target signatures, and backgrounds. They must also be able to perform many types of missions, often simultaneously.

In order to relate system performance with system cost, the sensor performance must be linked to specific system architectures with well-defined sensor payloads and associated spacecraft bus, communication, ground, and launch systems. The focus in this discussion is a key system element: the infrared-sensor payload configured for the detection and tracking of theater missiles. The sensor is assumed to be deployed in a geostationary orbit with a field of regard to cover the anticipated threat areas, for example, the Middle East and Northeast Asia.

Mean frequency of occurrence of clouds above 10 kilometers (University of Wisconsin HIRS-2 data, August 1989–1994). Earth's surface is never completely covered with clouds on any given day. High-altitude clouds are more likely to occur at lower latitudes. The analysis in this study generates the probability of missile warning when clouds of a given type and altitude are present at the locations of interest. A global cloud statistical model developed at Aerospace indicates that, based on six years of data, clouds at 10 kilometers or above occur over Northeast Asia about 20 to 30 percent of the time during the summer, with a maximum cloud coverage of 40 percent. In light of these statistics, sensor performance must be evaluated against clouds ranging in altitude up to 10 kilometers or possibly higher.

Typical mission requirements include the minimum detectable target, time of first report, size of the geographical area of interest, accuracy of the reported launch location, heading of the target, and accuracy of the predicted impact point. These are met with a set of properly sized infrared sensors that are configured with an optimized satellite-constellation architecture. The constellation is deployed in a geostationary orbit, which determines the number of satellites, data communication rates and other elements of the infrastructure, and space- and ground-based processing systems. Sensor sizing is driven by the required sensor sensitivity, spatial resolution at the target range, revisit time (time between looks) or target-update rate, and the selection of appropriate spectral bands for discriminating targets from backgrounds.

Once the system performance parameters are defined and the constellation architecture selected, an infrared sensor-system design is synthesized and the sensor's first-order technical design parameters are obtained:

sensor type (scanner or starer)
system field of regard and scan pattern
telescope field of view, detector pixel instantaneous field of view, aperture and focal length
system scan rate/staring duration
focal-plane definition (single or dual color), sensitivity, topology, and frame rate
signal-processing data rates and functional definition
overall system digital output data rates

The next level of synthesis fleshes out the infrared sensor subsystems in enough detail to allow reasonable estimates of the mass, power, and volume of the orbital components of the system. Meaningful technology risk assessments can now be formulated.

A variety of linked analysis software tools and databases execute the payload design and sizing process. For example, defining the focal plane includes selecting the focal-plane detector material and optical cut-off wavelength, spatial layout, detector or pixel dimensions, and readout rate(s). The sensor sensitivity requirement is translated into a focal-plane sensitivity constraint, which allows selection of the focal-plane temperature, using a thermal-noise model specifically tailored for the chosen detector material (e.g., mercury cadmium telluride). The focal-plane topology (total detector count) and maximum readout rate then allow determination of the electrical power, which, with the focal-plane temperature, serve as inputs to a model for determining the technology, size, power, and volume of the cryogenic cooling system.

The various simulation tools developed and used by Aerospace and their interconnectivity for the evaluation of system performance. Such simulation tools must be built to accurately model the interaction of the sensors with relevant target and background characteristics over all possible viewing areas. One of the main goals of the tool development is to incorporate the effects of "real" clutter phenomena, such as cloud edges and sun glints, within system-level analyses. A scene-based methodology generates appropriate clutter statistics, which are included within the constellation-level simulations.

A state-of-the-art optical design and tolerancing program uses the design parameters to select and refine a specific telescope optical design for the optical system. The optical program data are used to estimate the mass and volume of the optical subsystem and the power required to scan and point the line-of-sight mirror. The subsystem masses, power dissipations, and volumes are "rolled-up" into an overall payload mass, power, and configuration, which is then used to size the spacecraft bus and to select the launch system. The detailed subsystem parameters, along with the subsystem masses, power-consumption levels, and configurations, are then passed to an appropriate cost-estimating tool(s).

In order to relate system-level performance with cost, the foregoing process is carried out as a function of sensor performance parameters by varying, for example, the sensor noise-equivalent target, revisit time, and ground-sample distance (resolution). For such parametric analyses, scaling relationships are often used to interpolate subsystem mass, power, and volume estimates between design points. This is appropriate once a detailed design is developed for a basis; excursions from this fiducial design then use the appropriate scaling relationships.

Sensor Performance Simulation Tools

The remainder of the discussion focuses on the sensor taken to be an infrared scanner; constellation system-level simulation tools are used to calculate end-to-end performance. An example is provided wherein the linear size of the sensor ground sample (spatial resolution or detector "footprint") is systematically varied up to a factor of two to illustrate the impact of sensor susceptibility to background clutter. The level of this clutter is also varied over a wide range. Variable ground-sample size is used to derive corresponding sensor system designs from which payload masses are determined. To more clearly illustrate this sensitivity trade, cost is assumed to be related only to the infrared sensor mass, as a rough approximation. The specific cost-benefit relationships developed by this example are illustrated in the next section.

Solar-scatter geometry: As the sensor line of sight shifts from nadir to the limb, both the range to target and the path length through the atmosphere increase, while the minimum possible solar-scatter angle (SCA) decreases. When the sensor is viewing targets at the nadir (look-zenith angle [LZA] equals zero degrees), the minimum possible SCA is 90 degrees, whereas when viewing low-altitude targets at the limb (LZA equals 90 degrees), the sun can be directly in the sensor line of sight (i.e., SCA can be zero degrees). Unfortunately, most of the surface area covered by a given space sensor viewing Earth lies at the larger LZAs. Overlapping coverage provided by a constellation of many sensors mitigates this problem.

Simulating the performance of a space-based surveillance system involves the use of an ensemble of software models and databases:

SSGM—Synthetic Scene Generation Model: encapsulates many phenomenology codes under one architecture; developed by Photon Research Associates Inc. for the DOD Ballistic Missile Defense Organization
CLDSIM—cloud-scene simulation model incorporated in SSGM
VISTAS—Visible and Infrared Sensor Trades, Analyses, and Simulations model: combines classical image-processing techniques with detailed sensor models to produce static and time-dependent simulations of a variety of sensor systems, including imaging, tracking, and point-target-detection scanners and starers; designed and coded by Aerospace.
TRADIX—constellation-level analysis tool that combines electro-optical sensor models with target, background, and atmospheric models to evaluate system performance; developed by Aerospace.

The simulation begins with the generation by SSGM of a set of shortwave infrared Earth-cloud background scenes. Scenes are selected from the database of weather satellite imagery. The images are pixelized, and the infrared properties of the scene elements are inserted into the image database. A key step in this process is the use of CLDSIM to simulate the solar scatter from cloud tops at various altitudes. Scenes with only terrain, sea surfaces, and low-altitude clouds usually do not generate much clutter in the chosen spectral band. Solar reflections from high-altitude clouds, on the other hand, can cause a high degree of clutter, which can, in turn, stress the sensor's ability to detect targets of interest. SSGM can generate selected atmospheric properties, a specified spatial resolution, and a matrix of scenes with a variety of viewing geometries in the desired spectral band.

For the evaluation of sensor performance, VISTAS models the imaging chain of the electro-optical sensor, from the background scene input to the signal-processor output. The sensor-system transfer function is applied to a high-resolution input scene, such as those produced by SSGM. The transfer function (output vs. input) includes the effects of the optical point spread function, that is, the blur, and for a scanner, the temporal aperture caused by the scan motion during the integration time. The blurred scenes are resampled at the system resolution, and clutter-rejection filters are applied. The output is calibrated to account for the sensor system's response to the target intensity. The output scenes are then analyzed for figures of merit, such as the standard deviation of the sensor response, which are statistical representations of the clutter of the background scene and sensor noise for the sensor design under consideration.

TRADIX models space sensors operating in both above- and below-the-horizon modes, from the visible to the long-wave infrared. It contains a model for the infrared signature of the missile body, plume intensity and trajectory profiles, clutter statistics generated by SSGM/VISTAS, background models including straylight from nonrejected earthshine and sunshine, and the atmospheric path radiance and transmission. These models are integrated with the Aerospace orbit propagation library, ASTROLIB, to provide a dynamic simulation tool for studying the constellation-wide performance of electro-optical sensors.

Nominal scene, left, used for clutter performance analysis: clouds at 2–8-kilometer altitude, 200-meter resolution, and midlatitude summer atmosphere. The database is referred to as nominal because it contains mostly low- to mid-altitude clouds and results in moderate clutter levels, depending on the sensor design and solar angle. Also, these clouds occur quite frequently over a wide range of latitudes. The brighter clouds are at higher altitudes where there is less attenuation of the sunlight. Stressing scene, right, used for clutter performance analysis: clouds at 4–10-kilometer altitude, 200-meter resolution, and midlatitude summer atmosphere. The database is referred to as stressing because it contains cirrus ice clouds at 10-kilometer altitude and can generate quite severe clutter levels. These clouds do occur less frequently than those contained in the nominal scene.

Critical inputs to the above tools include focal-plane pixel topology, sensor noise characteristics, details of the filters and signal processing, optical design and straylight rejection capability, platform drift and jitter, constellation orbits and phasing, and concepts of operation, that is, scan modes, revisit times, and others.

Information on the frequency of occurrence of meteorological conditions is critical to the performance of a space-based infrared surveillance system against Earth backgrounds. Aerospace uses a global cloud statistical model that is based on the University of Wisconsin HIRS-2 (High-Resolution Infrared Sounder) data from the National Oceanographic and Atmospheric Administration polar orbiters. The database incorporated in this model provides a basis for assessing system performance against clouds of various altitudes for a given time of year at a specified geographical location. Typical displays of such data would be shown on a world projection in terms of the maximum, minimum, and mean distributions of the probability of clouds above a given altitude during a specified month.

Scene-Based Clutter Analysis

Probability of false exceedance plotted as a function of threshold intensity for infrared scanners viewing the stressing background scene. A threshold of approximately 6 kilowatts per steradian for the 1.8-kilometer-footprint scanner design corresponds to a false exceedance of 10^-4. When that same threshold is applied to the 3.6-kilometer-footprint scanner design, the false exceedance is approximately three orders of magnitude higher (10^-1), whereas the reduction in the number of detector channels is only a factor of four.

The performance of an Earth-viewing infrared sensor, designed to operate in the shortwave infrared atmospheric-absorption band to block signals from terrestrial sources, is dominated by the structure in the background scene. This structure is caused predominantly by sunlight reflected from clouds in the scene and depends on the sensor-cloud sun-viewing geometry, represented by the cloud "look-zenith angle" and the "solar-scatter angle." The solar-scatter angle, which determines the intensity of solar scatter off the cloud tops, is dominant; very small values of the solar-scatter angle result in very intense scattering. The other angle, the look-zenith angle, defines the projection of the SSGM cloud scene in the sensor line of sight, in reference to the range from the sensor to the cloud tops. The look-zenith angle is also directly related to the path length through the atmosphere to a target at a given altitude, and hence to the target's apparent irradiance for a given time after launch. At high-zenith angles, low-altitude targets viewed within the Earth limb suffer from the worst combination of range, atmospheric transmission loss, and solar-induced background clutter. Unfortunately, most of the Earth's surface viewed by a sensor in space lies at the larger look-zenith angles. However, the overlapping coverage of multiple sensors can be used to mitigate this problem.

The two SSGM scenes selected in this study represent a "nominal" case containing low- to mid-altitude clouds, and a "stressing" case with mid- to high-altitude clouds. The terms nominal and stressing refer to the level of the clutter generated when the scene is passed through a typical sensor simulation. Each scene covers an area of 512 X 1170 kilometers at nadir with a 200-meter spatial resolution, as seen through a midlatitude summer atmosphere at a look-zenith angle of 60 degrees and solar-scatter angle of 90 degrees; geometric projection effects at a look-zenith angle of 60 degrees shorten the apparent size of these scenes to 510 X 570 kilometers. The brighter clouds in these images are at higher altitudes where there is less attenuation of sunlight both before and after it scatters from the cloud tops.

1.8-kilometer footprint

3.6-kilometer footprint

Infrared-scanner outputs in the simulation for the stressing background scene. For a scanner, each background input scene results in a single simulated static output scene. The intensity of a target relative to the surrounding background is the principal method for detection. Gray represents the mean level of zero, while white and black represent positive and negative exceedances (false indications of a target in the field of view). In order to provide a visual guide to the clutter levels in these two output scenes, four "target" markers with an intensity of 5 kilowatts per steradian are imbedded in each scene. All four are clearly visible in the 1.8-kilometer case, whereas only the central, isolated target is visible in the 3.6-kilometer scene. These output scenes are used to provide a statistical representation of the clutter in terms of the number of exceedances vs. threshold level for the background scene and sensor design.

1.8-kilometer footprint

3.6-kilometer footprint

Infrared-scanner simulation: outputs thresholded at 6 kilowatts per steradian for the stressing background scene. If the 6-kilowatts-per-steradian threshold is applied to the output scenes, the impact of sensor footprint on clutter response is immediately apparent: clutter is reduced in both cases, but there is a more pronounced effect for the 1.8-kilometer footprint. Threshold levels are usually set by the limitations of the onboard processor or the ground communications link. The need to limit the number of false alarms reported by the system can also be a significant constraint. False-exceedance levels of approximately 10^-4 to 10^-3 are typical.

The most significant caveat concerning this analysis is the use of the CLDSIM model within SSGM. A number of uncertainties are inherent in this model, one being the model for the solar scattering from the clouds. Using detailed comparisons with actual space sensor data from the MSTI-3 (Miniature Sensor Technology Integration) sensor that collected information on Earth and Earth-limb clutter, the resultant uncertainty in the apparent cloud brightness has been estimated to be certainly less than a factor of three. The impact of this uncertainty was addressed by scaling the intensity of each scene, in the nominal case by one-third and one-half and in the stressing case by two and three, thus providing a total of six cases. In this way the effect on system performance of varying the cloud types and altitudes and the additional impact of the SSGM modeling uncertainties were quantified.

In viewing the Earth-cloud background scenes, a scanning sensor typically responds only to changes in the signal, in effect performing a subtraction of the mean background radiance, thus providing an indication of the target intensity above the mean. The effectiveness of this suppression of clutter caused by radiance differences in the scene depends on the instantaneous field of view of the detectors, the pixel footprint. In this study, scanning-sensor designs with footprints of 1.8, 2.6, and 3.6 kilometers were used at a nominal range of 40,000 kilometers (corresponding to a look-zenith angle of approximately 75 degrees for a geostationary satellite). The output scenes then exhibit a mean background level of zero, with positive and negative values apparent at the cloud edges. These simulation outputs were used to determine the probabilities of false exceedance (false indication of a target in the field of view) versus intensity threshold for the background clutter.

shortwave IR intensity of theater missile

The apparent shortwave-infrared intensity of a hypothetical theater missile vs. time after launch when viewed at nadir and the limb; includes attenuation by the atmosphere.

In order to set a threshold, an acceptable false exceedance rate for each sensor design must be determined. For architectures where the mission data are processed on the ground, this involves several factors: the number of detectors in the focal plane, the sampling rate, the number of bytes per sample, and the capacity of the downlink, as well as some knowledge of the target detection algorithm. For example, a scanner design with a 1.8-kilometer footprint could indicate a threshold of 6 kilowatts per steradian for a false exceedance rate per pixel of 10^-4, whereas a design for a 3.6-kilometer footprint viewing the same scene would indicate a false exceedance rate per pixel of 10^-1 for the same threshold intensity. Of course the smaller footprint design would require many more detectors and a larger telescope, hence a heavier payload and greater cost. The ground footprint of an infrared scanner is a key parameter in determining sensor performance against structured backgrounds.

In a more complete analysis, one must account for the sensor noise arising from the natural fluctuations in the scene and from the electronics, usually expressed as a noise-equivalent target intensity. This accounting is expressed in a threshold-versus-exceedance distribution. If the clutter noise from the background is random in character, it can be expressed as a clutter- equivalent target intensity; then a system-equivalent target intensity can be determined simply as the square root of the sum of the squares. However, clutter from clouds in natural background scenes is usually far from random, so that more complex methods are required for combining the sensor noise with the background clutter distribution.

Target Response

instantaneous line of sight coverage of 2 theaters

Instantaneous line-of-sight coverage of two theaters by five (four pinched) geostationary satellites. The 3-X3-degree target grid results in approximately 70 launch locations per theater. A representative type of theater missile has a burnout time of approximately 60 seconds after launch. The ability to optimize satellite locations and thus trade between theater performance and global coverage is most easily achieved with geostationary constellations. The longitudes of the five geostationary satellites are shown here on a cylindrical projection of Earth, along with the boundaries of the two theaters and the constellation's line-of-sight coverage. Essentially 100-percent triple coverage is provided over both theaters of operation.

Target response is a scale factor that describes the attenuation of a target through the sensor system. The response of a scanning sensor to an unresolved target, that is, a "point source," depends on several factors. These are the blurring caused by the optics, the temporal aperture caused by the scan motion during the integration time, the sampling of the blurred target by the focal plane, the target phasing (i.e., the location of the target relative to the center of a pixel), and the electronic filtering.

For a fast-scanning sensor system, the response does not depend on the temporal characteristics of the target so the calculation is fairly straightforward. The target response can be evaluated by constructing a scene in which a grid of many point sources (usually 1 kilowatt per steradian) are spaced far enough apart to avoid interference, each offset randomly by a small amount to make the grid nonuniform, thus to ensure many different target phasings. This target grid is then passed through the same simulation process as the background scenes, namely blurring, downsampling, and filtering. The peak response from each target is determined, and the average is taken as the mean target response. All simulated clutter scenes are divided by this target response so they can be referenced to apparent intensities at the sensor aperture.

Constellation Performance

To evaluate the performance of a space-based infrared surveillance architecture against a variety of target and background conditions, the sensor response to both must be combined in a constellation-level simulation. This is done by first choosing one of the cloud databases contained within the SSGM, then generating scenes spanning the entire range of viewing geometries and sun angles for the selected sensor constellation. The scenes are then processed through the detailed sensor simulation tool, VISTAS, to produce a set of false-exceedance-threshold clutter distributions. These are combined with the sensor noise-equivalent target intensity in TRADIX, and the threshold and minimum detectable target are calculated as functions of the look-zenith and solar-scatter angles, for the required probability of detection and false exceedance.

Theater-missile-warning stereo performance of five (four pinched) satellites against highly stressing clutter. The stereo performance results across the entire Eurasian landmass are shown for all three sensor designs against the highly stressing clutter level. The performance for the Middle East is indeed very close to that for Northeast Asia. These designs would be even further stressed if the system were being asked to perform additional missions simultaneously.

Within TRADIX, the targets and satellites are propagated in Earth-central-inertial coordinates with an appropriate sampling interval, typically 5 to 15 minutes. This is carried out over a period of a day at various times during the year to explore the effects of seasonal variations in the sun's latitude. For a constellation in geostationary orbits and target launch sites in the northern hemisphere, the most stressing solar-scattering angles are close to the summer solstice, and the resultant background clutter is the dominant effect in the system performance. On the other hand, the effect of solar straylight can be more stressing close to the solar equinoxes.

Each sensor-target line of sight at each time results in a look-zenith-angle/solar-scatter-angle pair, which, with the clutter data, sets a minimum-detectable-target threshold for that sensor, and a time of first detection. For example, three out of four detections produces a "3of4" report from that sensor, and two such reports by separate sensors result in a stereo report for the constellation. Target-detection and report-time statistics are thus generated for each sensor design against each structured background for the mission of interest. This procedure has been applied to missions of global-and theater-missile warning.

Theater-missile-warning stereo performance sensitivity to sensor footprint and clutter. Stereo tracking is necessary if stressing requirements associated with launch point determination and impact point prediction are to be met. The impact of the larger footprints on the stereo performance of the three sensor designs is severe. A 95-percent probability of stereo detection before burnout is only achievable for the 2.6-kilometer or better design against the stressing clutter level. The 3.6-kilometer design can only meet a stereo probability of detection of 95 percent at the nominal clutter level or below. Ten-kilometer clouds can be expected over Northeast Asia approximately 20 to 30 percent of the time during the summer months, and six-kilometer clouds can be expected 40 to 50 percent of that time.

For a typical global analysis, target launch sites are uniformly spread over the surface of Earth from –90- to +90-degree latitude. A target spatial pattern with a resolution of 3 X 3 degrees (Earth central angle) generates 4586 distinct target launch locations that represent equal-sized areas on the surface of Earth. Missiles are usually launched in 12- to 36-azimuthal directions in order to allow for aspect angle effects on the apparent booster signature. The point of such an analysis is to obtain a measure of system performance that is not scenario driven. As many as 50 million target launches may be run in order to determine global performance for a particular space-surveillance constellation.

For the theater-missile-warning mission, a representative short-range missile was assumed with an infrared intensity profile increasing as the missile rose through the atmosphere to a maximum, then decreasing as the afterburning of the exhaust diminished (as viewed from a broadside aspect at the two extreme look-zenith angles of 0 and 90 degrees, corresponding to the nadir and the limb). The analysis was limited to the worst-case epoch (date or day of year) for clutter-limited detection and was focused on two theaters of operation, one in the Middle East, the other in Northeast Asia. A constellation of five satellites, with four of the satellites "pinched" to cover the Eurasian landmass, was selected to provide excellent overlapping coverage of both areas simultaneously.

A short-wave-infrared line scanner was selected as the sensor-design option. Although not the most effective choice for clutter suppression, a scanner reduces the program risks associated with extreme line-of-sight stability and focal-plane producibility associated with infrared-starer designs. On the other hand, for acceptable performance against highly structured backgrounds, it is necessary that a scanner have a relatively small footprint. In this study, three designs with footprints of 1.8, 2.6, and 3.6 kilometers were considered.

Performance sensitivity to sensor footprint and payload mass at the various shortwave-infrared clutter levels with line-of-sight stereo coverage. Essentially, the graph shows the cost of designing an infrared sensor with guaranteed performance against an uncertain level of background clutter. The larger footprint designs are less robust against increasing clutter levels. The data show that while there are significant cost savings to be made in lower payload mass (and power), the performance penalty in stereo track capability associated with a larger footprint design may be severe.

The infrared-scanner optics consisted of a triplet refractor and a two-axis entry flat for scanning. For a low noise-equivalent-target intensity, sensor aperture was traded against time delay and integration capability on the focal plane. The result was a 27-centimeter aperture with 12 stages of time delay and integration. A 2-second revisit time taken as a mission requirement necessitated a scan rate of 4 degrees per second and resulted in a noise-equivalent-target intensity of about 1 kilowatt per steradian at 40,000 kilometers. A single-hit probability of detection of 95 percent was chosen; this led to a cumulative probability of detection of 99 percent for the 3of4-hits algorithm. The resultant probability of detection and false exceedance combination led to a minimum-detectable target as a function of viewing geometry for each sensor design, background scene, and clutter scale factor. The TRADIX constellation analysis tool was then used to evaluate the performance of the three sensor designs against the nominal and stressing backgrounds.

Conclusion

The costs associated with the deployment of a surveillance satellite are closely related to the payload mass, both in regard to the payload itself and that of the launch vehicle required to lift it and its satellite platform into a geostationary orbit. Accordingly, rather than attempting to carry out a detailed cost analysis, this study was focused on estimations of payload mass, which included the telescope, sensor housing, and scan mirror; the focal-plane assemblies and signal processors; power supplies; and other subsystem masses. The highest-performing sensor with the smallest footprint and best clutter suppression ends up also being the heaviest and most expensive. The overall costs of this highest-performing system would have to be weighed against the value of the mission to the national interest; such considerations are beyond the scope of this investigation.

The study described here illustrates the application of analytical tools and databases designed and assembled at Aerospace to support the development of advanced space-based surveillance systems. The example shown represents a hypothetical system for the specific purpose of missile warning that covers two widely separated geographical areas of concern. Aerospace is currently applying comparable analyses in its support to the Air Force Space and Missile Systems Center and the Ballistic Missile Defense Organization for the development of both SBIR High and Low systems.