SITIS Archives - Topic Details
Program:  SBIR
Topic Num:  AF071-159 (AirForce)
Title:  Real-Time Active Sensor Target Scene Generator
Research & Technical Areas:  Weapons

 The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), which controls the export and import of defense-related material and services. Offerors must disclose any proposed use of foreign nationals, their country of origin, and what tasks each would accomplish in the statement of work in accordance with section 3.5.b.(7) of the solicitation.
  Objective:  Develop Innovative Low Cost Technique for Generating Major RF Scatterers in a Simulated Scene including Real-Time Updating their Range, Velocity, and Radar Cross Section
  Description:  Target detection and classification through the use of active imaging techniques is required for the next generation of short-range, high-resolution, high-speed fuzes. Active systems provide a high probability of detection and excellent countermeasure resistance against advanced surface and air targets. Unlike previous short-range sensors architectures, modern concepts are software driven programmable devices that use sophisticated radar signal processing techniques. The robustness of these new designs now exceeds the capability to thoroughly evaluate their operating envelope using a modern multi-channel RF hardware-in-the-loop target simulator. The issue is that the existing target simulator architectures can not affordably generate enough scatterers in the scene to adequately represent operational scenes. Although operating frequencies of these active sensors could range from 1-800 GHz, the key to this effort is manipulating the down-converted intermediate frequency to create the multiple scatterers needed and then updating the characteristics real-time as the sensor moves through the scene. For purposes of this effort, the problem can be partitioned into two levels of performance. The first level would be a 2-D rendering of a scene. This 2-D scene is defined by the plane normal to the surface of the earth and intersecting the velocity vector of the sensor flying through the scene. The 2-D scene rendering could also be called the ground profiling mode. The second level of performance would be a 3-D rendering of the scene. This would include a flat earth rendering along with man made objects with enough detail to support evaluation of target detection and classification algorithms for short range active imaging sensors. The 3-D scene rendering could also be called the active imaging mode. As a starting point assume the ground profiling mode starts an altitude no greater than 3 kilometers and the active imaging mode starts at an altitude no greater than 100 meters. Both would be falling to the earth at depression angles from 10 to 90 degrees and velocities from 25-500 meters per second. The ideal implementation of this concept would be non-cooperative in that the target scene generator would operate autonomously with respect to the short range sensor for which it is generating the scene. The sensor’s transmitted energy would be radiated or injected into the simulator where it would be manipulated to add the appropriate scene and then radiated or injected back into the RF sensor under test.

  PHASE I: The Phase I effort will be a thorough concept refinement and analysis to show as a minimum a realistic 2-D scene generating concept with a development path to the 3-D performance.
  
  PHASE II: The Phase II effort will demonstrate feasibility of the chosen concept.

  DUAL USE COMMERCIALIZATION: Military application: Modern software driven sensor developments would use this robust, dynamic hardware and algorithm testing environment as a software validation tool prior to live drop testing of the assets. Commercial application: This technology will be poised to provide the only affordable solution for realistic scene generation to aid in hardware and software development for specific sensor applications.

  References:  1. “Introduction to Airborne Radar”, G. W. Stimson, SciTech Publishing, Inc. 1998. 2. “An All-Digital Image Synthesizer for Countering High-Resolution Imaging Radars”, R. T. Ekestorm and C. Karow, Nava Postgraduate School Thesis, September 2000; AD# ADA381835. 3. “A Comparison of DDS and DRFM Techniques in Generation of ‘Smart Noise’ Jamming Waveforms”, C. J. Watson, Nava Postgraduate School Thesis, September 1996; AD# ADA324218. 4. “Fuze Air-to-Surface Technology (FAST) Program”, Program Research and Development Announcement (PRDA), www.fedbizopps.gov , 2003.

Keywords:  Active Imaging, Fuzing, Scene Generation, Active Sensors, Digital RF Memory, RF Target Simulator

Questions and Answers:
Q: 1. Can you describe the radar backscattering model to be used here? Does the reference to “Major RF Scatterers” in the Objective indicate that the terrain is to be modeled as an array of point scatterers? Or are the scatterers inherently directional (e.g., facet-based)? Will discrete scatterers be used to model man-made objects, but the return from natural surfaces produced using an area scattering model, with a radar cross section density which is a function of terrain type, grazing angle, frequency, polarization? If the latter, would the RCS be defined by a Weibull PDF?

2. If point scatterers, how many to be simulated? For simplicity, are only scatterers in the antenna mainlobe to be included in the return, or everything in the scene weighted by the full antenna pattern?

3. Can you say anything about the radar and antenna parameters? Obviously radar bandwidth is key to sizing the scene generator.

4. How much latency will be allowed in the simulator?

5. Can you give a more complete description of 2D mode? In this mode is the radar CW with spatial resolution provided only by an electronically steered antenna?

6. Finally, is the scattering model to be used in this topic different from that of topic AF071-165?
A: 1. For Topic 159, each of the scatterers is represented in hardware by a specific range delay, range rate (Doppler or velocity), and radar cross section (returned power level). The backscattering model used is defined by the user of the target simulator. The emphasis of this topic is for non-cooperative receiving of a radar sensor's transmitted energy, modifying the received energy to add the characteristics of each of the scatterers, and re-transmitting the modified signal back to the radar in real-time. The sub-questions you have relate more to a entirely synthetic radar simulator and scene generation environment.

2. The 2-D scene rendering would have 2^N scatterers generated real time with a goal of 1024 simultaneous scatterers. The 3-D version of the simulator hardware is envisioned as multiple parallel 2-D channels that are coherently and simultaneously updated.

3. The goal for the instantaneous bandwith of the simulator is to be greater than or equal to 500 MHz. The radars antenna pattern and modulation are not important as long as its bandwidth doesn't exceed that of the simulator.

4. It is envisioned that the simulator will be operated as a continuous streaming of analog signals into and out of the system. Given that delays through the system could add 100's of nanoseconds to the minimum delay, there must be enough memory available in the concept to delay transmission until the radars next coherent processing interval starts. Our current radar target simulator has ~ 8 milliseconds of memory which has met all of our needs to date.

5. Takes the case of a radar approaching the ground at a 45 degree angle and project its velocity vector until it touches the ground. Next, pass a plane through the velocity vector and intersect the ground. The line formed where the plane intersects the ground defines the region of interest for the 2-D rendering and is broken into the 2^N scatterers with a resolution and window around the velocity vector to be defined by the user. Lastly, each of these segments along the ground would be characterized by one set of range, velocity, and radar cross section values. If one assumes 1024 segments on the ground, then there would be 1024 scatterers generated that are mathematically related and simultaneouly updated at each simulation time step.

6. Topic 165 is a totally synthetic simulation with orders of magnitude more target and scene fidelity than topic 159. It also addresses all of the issues found recently when simulation of multiple coherent receivers was needed. One could envision a time when a flyout scene from topic 165 could be decimated and used as the hardware input control file for topic 159, but we are far from that until each of these capabilities is developed and matured.
Q: Following up on items 4 and 5:

4) Just to rephrase since this is an important point - The scene generator is allowed to assume that each coherent burst is the same – including the antenna pattern - and delay its reponse by one burst? Therefore easing the design difficulty by permitting milliseconds of latency rather than nanoseconds. Correct?

5a) Sorry, it is difficult to be sure I have a correct understanding of 2D mode without a figure, so please bear with me. From #5, regarding “Next, pass a plane through the velocity vector and intersect the ground” - by a plane “through” the velocity vector you mean a plane “normal” to the velocity vector? And therefore the intersection with the ground is a line, extending in the cross-range direction, which is perpendicular to the velocity vector? This implies that there will be a large swath of scatterers clustered around the intersection of the velocity vector with the ground with little differential Doppler (assuming immobile scatterers). However, if you move far enough in cross-range along the line, Doppler will decrease quadratically (at least at first). Is this correct?

5b) “Lastly, each of these segments along the ground would be characterized by one set of range, velocity, and radar cross section values. If one assumes 1024 segments on the ground, then there would be 1024 scatterers generated that are mathematically related and simultaneouly updated at each simulation time step.”

Is it a correct interpretatioin that the reference to scatterer “velocity” is not simply the relative velocity of a fixed scatterer to the moving sensor? Your intent is that the scatterers may be mobile, each with a unique velocity vector?

5c) The 5b quote also answered my question as to whether it was the task of the scene generator to determine how scatterer cross section evolves with time based on the scene geometry. From your answer it appears it is not – the scene generator is simply given a new value each “simulation time step.” Correct? Approximately how long will each time step be?
A: 4) Yes, you can assume that the next coherent burst is the same as the first for performing second-time-around simulation.

5a) The plane passes through the velocity vector such that it is parallel, not perpendicular to the velocity vector. This ground profile is basically the projection of the velocity vector onto the surface of the earth.

5b) For our purposes, we assume everything on the ground is stationary, but it is really up to the user defined simulation control software to include effects of moving portions of the scene during the simulation.

5c) That is correct. A new array of values would be given to the simulator at each time step. Currently our deterministic update rate is 1 kHz with plans for 10 kHz deterministic operation. Alternatively to real-time control, a complete 3 minute flyout of pre-processed control words could be loaded into the simulator and the simulation started on command. The fact that each flyout only lasts seconds to minutes could be included in the design considerations of the system architecture.
Q: 1. Can you describe the radar backscattering model to be used here? Does the reference to “Major RF Scatterers” in the Objective indicate that the terrain is to be modeled as an array of point scatterers? Or are the scatterers inherently directional (e.g., facet-based)? Will discrete scatterers be used to model man-made objects, but the return from natural surfaces produced using an area scattering model, with a radar cross section density which is a function of terrain type, grazing angle, frequency, polarization? If the latter, would the RCS be defined by a Weibull PDF?

2. If point scatterers, how many to be simulated? For simplicity, are only scatterers in the antenna mainlobe to be included in the return, or everything in the scene weighted by the full antenna pattern?

3. Can you say anything about the radar and antenna parameters? Obviously radar bandwidth is key to sizing the scene generator.

4. How much latency will be allowed in the simulator?

5. Can you give a more complete description of 2D mode? In this mode is the radar CW with spatial resolution provided only by an electronically steered antenna?

6. Finally, is the scattering model to be used in this topic different from that of topic AF071-165?
A: 1. For Topic 159, each of the scatterers is represented in hardware by a specific range delay, range rate (Doppler or velocity), and radar cross section (returned power level). The backscattering model used is defined by the user of the target simulator. The emphasis of this topic is for non-cooperative receiving of a radar sensor's transmitted energy, modifying the received energy to add the characteristics of each of the scatterers, and re-transmitting the modified signal back to the radar in real-time. The sub-questions you have relate more to a entirely synthetic radar simulator and scene generation environment.

2. The 2-D scene rendering would have 2^N scatterers generated real time with a goal of 1024 simultaneous scatterers. The 3-D version of the simulator hardware is envisioned as multiple parallel 2-D channels that are coherently and simultaneously updated.

3. The goal for the instantaneous bandwith of the simulator is to be greater than or equal to 500 MHz. The radars antenna pattern and modulation are not important as long as its bandwidth doesn't exceed that of the simulator.

4. It is envisioned that the simulator will be operated as a continuous streaming of analog signals into and out of the system. Given that delays through the system could add 100's of nanoseconds to the minimum delay, there must be enough memory available in the concept to delay transmission until the radars next coherent processing interval starts. Our current radar target simulator has ~ 8 milliseconds of memory which has met all of our needs to date.

5. Takes the case of a radar approaching the ground at a 45 degree angle and project its velocity vector until it touches the ground. Next, pass a plane through the velocity vector and intersect the ground. The line formed where the plane intersects the ground defines the region of interest for the 2-D rendering and is broken into the 2^N scatterers with a resolution and window around the velocity vector to be defined by the user. Lastly, each of these segments along the ground would be characterized by one set of range, velocity, and radar cross section values. If one assumes 1024 segments on the ground, then there would be 1024 scatterers generated that are mathematically related and simultaneouly updated at each simulation time step.

6. Topic 165 is a totally synthetic simulation with orders of magnitude more target and scene fidelity than topic 159. It also addresses all of the issues found recently when simulation of multiple coherent receivers was needed. One could envision a time when a flyout scene from topic 165 could be decimated and used as the hardware input control file for topic 159, but we are far from that until each of these capabilities is developed and matured.
Q: Following up on items 4 and 5:

4) Just to rephrase since this is an important point - The scene generator is allowed to assume that each coherent burst is the same – including the antenna pattern - and delay its reponse by one burst? Therefore easing the design difficulty by permitting milliseconds of latency rather than nanoseconds. Correct?

5a) Sorry, it is difficult to be sure I have a correct understanding of 2D mode without a figure, so please bear with me. From #5, regarding “Next, pass a plane through the velocity vector and intersect the ground” - by a plane “through” the velocity vector you mean a plane “normal” to the velocity vector? And therefore the intersection with the ground is a line, extending in the cross-range direction, which is perpendicular to the velocity vector? This implies that there will be a large swath of scatterers clustered around the intersection of the velocity vector with the ground with little differential Doppler (assuming immobile scatterers). However, if you move far enough in cross-range along the line, Doppler will decrease quadratically (at least at first). Is this correct?

5b) “Lastly, each of these segments along the ground would be characterized by one set of range, velocity, and radar cross section values. If one assumes 1024 segments on the ground, then there would be 1024 scatterers generated that are mathematically related and simultaneouly updated at each simulation time step.”

Is it a correct interpretatioin that the reference to scatterer “velocity” is not simply the relative velocity of a fixed scatterer to the moving sensor? Your intent is that the scatterers may be mobile, each with a unique velocity vector?

5c) The 5b quote also answered my question as to whether it was the task of the scene generator to determine how scatterer cross section evolves with time based on the scene geometry. From your answer it appears it is not – the scene generator is simply given a new value each “simulation time step.” Correct? Approximately how long will each time step be?
A: 4) Yes, you can assume that the next coherent burst is the same as the first for performing second-time-around simulation.

5a) The plane passes through the velocity vector such that it is parallel, not perpendicular to the velocity vector. This ground profile is basically the projection of the velocity vector onto the surface of the earth.

5b) For our purposes, we assume everything on the ground is stationary, but it is really up to the user defined simulation control software to include effects of moving portions of the scene during the simulation.

5c) That is correct. A new array of values would be given to the simulator at each time step. Currently our deterministic update rate is 1 kHz with plans for 10 kHz deterministic operation. Alternatively to real-time control, a complete 3 minute flyout of pre-processed control words could be loaded into the simulator and the simulation started on command. The fact that each flyout only lasts seconds to minutes could be included in the design considerations of the system architecture.

Record: of