Sensor Systems
Signal Processing
The Signal Processing Group at Adaptive Methods provides real-time solutions for many of the complex SONAR challenges that exist in today's U.S. Navy. We are a leader in providing low-cost/high-performance products that use latest technologies and employ flexible and practical designs. Adaptive Methods has long supported the COTS insertion programs and open-architecture initiatives. We will continue to be an advocate for these programs, striving to keep the development and maintenance costs low for our customers while maintaining the highest level of performance.
Core Developments:
GMP – Generic Multi-Processor System
The GMP system is an engineering software platform used for development and testing of real-time signal processing applications. The software has evolved over many years, in support of numerous Navy SONAR programs. Originally designed as an advanced rapid-prototyping platform, the system now forms the core of most real-time signal processing systems produced by Adaptive Methods. The system is specifically designed for multi-dimensional array applications. Engineers at Adaptive Methods can easily transition new functionality into the GMP architecture, while relying on existing functions (FFTs, etc.) to complete the required signal processing flow. The processing functions are easily manipulated using the Graphical Signal Processing Toolkit. Upon successful integration into the GMP system, the desired functionality and supporting infrastructure is extracted from the baseline for use in the delivered product.
Graphical Signal Processing Toolkit
A core part of the GMP system is the signal processing graphical design toolkit. This toolkit includes a Java-based processing network editor that allows the designer to manage a visual representation of the processing functions to be applied to the various input signals available to our signal processing products. The editor performs automatic layout of the processing graph, and full validation of all user input. Processing graphs are saved in an XML format for other tools to use in generating processing instructions for signal processing applications.
As new processing functions are introduced into GMP, engineers can easily extend the editor by providing specification of the functions and their parameters in XML files, supplemented by a small amount of function-specific Java code.
GMLIPC – Generic Middleware Layer Inter-Process Communication
GMLIPC is Adaptive Methods’ communications infrastructure, providing a unified application programming interface to all of the data messaging transports used by our products. The publish/subscribe API supports a high- performance proprietary transport used within Adaptive Methods’ distributed processing subsystems as well as customer and integrator-specified external transports, such as CORBA.
Sensor Acquisition & Recording
Adaptive Methods’ Sensor Integration Group develops front-end interfaces to collect and format data from physical sensors and legacy data busses for the U.S. Navy. Both custom and commercial off-the-shelf (COTS) hardware are used to tap the data sources and bring the raw information into symmetric multi-processor (SMP) servers for formatting.
The Sensor Integration Group designs and builds interfaces to a variety of sources including analog sensors, parallel and serial digital data busses, legacy NTDS systems, ATM, FDDI, and Ethernet networks. For many applications, Adaptive Methods designs custom hardware and integrates it with available COTS hardware. In some cases, Adaptive Methods consults with manufacturers on design of new COTS hardware to meet specific requirements.
The Sensor Integration Group draws on many disciplines including digital and analog hardware design, software engineering, signal processing and analysis, operating systems, networking, and communications theory.
SOFTWARE
One core strength of the Sensor Integration Group is the common software product, Multi-Input Data Acquisition System (MIDAS), which can simultaneously communicate with each interface. The software leverages the symmetric multi-processor capabilities and is highly multithreaded -- thus capable of handling several I/O and processing tasks simultaneously. Multiple tasks can run independent of one another or coordinate for increased acquisition capacity or parallel processing. The software can be synchronized across multiple servers for additional capacity or remote acquisition capabilities. MIDAS is supported under both the Windows and Linux platforms.
The PC Test Set 2000 is an engineering software tool for the Windows platform designed to acquire and process multi-channel data and provide display and analysis tools. Designed for use in conjunction with acoustic sonar systems, the PC Test Set acquires raw data from a telemetry source and transforms it into the time, spectral, or spatial domains. The data is calibrated and displayed for the user who can use the built-in tools to analyze the data. In addition, the PC Test Set has automated tools to inform technicians of the health of the sonar system.
HARDWARE
Distributed, multi-channel synchronized Analog-to-Digital converter system.
This flexible system can theoretically sample analog data simultaneously from an unlimited number of channels. Multiple PCI-based ADC cards are spread across multiple servers. Adaptive Methods consulted with a COTS PCI manufacturer, General Standards, to develop one of the leading PCI-based 24-bit acquisition boards on the market. The system has been implemented to acquire 576 channels of data from a SQS-53C hull mounted sonar array for the U.S. Navy, as well as other smaller applications.
Distributed, multiple bus digital receiver
Adaptive Methods developed the capability to receive data from an arbitrarily large digital data bus and applied it to tap four 20-bit busses from a legacy Navy ASW sonar processor.
We designed and built a custom interface to synchronize four PCI-based digital acquisition cards to a common clock and frame format. Due to the distributed, multi-tasking design, the system is able to receive and process data at up to 96MB/sec.
Serial acquisition with arbitrary frame format
Adaptive Methods developed an interface to a PCI-based serial capture card to acquire data from a serial data stream. Frame synchronization is handled entirely in software eliminating the need for any custom hardware for specific applications. The system was deployed to receive towed-sonar array telemetry for the U.S. Navy.
Advanced Development
Adaptive Methods has performed advanced development work for the U.S. Navy and prime contractors since its incorporation in 1973. Our advanced development goals are to transition new signal and information processing technologies to fleet operational systems. Our products are comprised mainly of sonar signal processing algorithms such as conventional and adaptive beamformers for passive and active applications, detection and classification algorithms, tactical decision aids, sensor-array shape estimation algorithms, and multi-sensor data fusion algorithms. In addition, the Advanced Development Group provides systems engineering services, hydrodynamic modeling, tow body design and fabrication, and test & evaluation support to the U.S. Navy and prime contractors.
Signal Processing Developments
Adaptive Methods’ advanced development algorithms and technologies are integrated into broadly applicable software applications that support a variety of Navy programs. Many of these algorithms are developed in a peer-review environment, providing an opportunity for our developers to interact with leading experts in the field.
Adaptive Methods has developed & implemented a variety of conventional beamformer (CBF) algorithms with wide applications in uniformly spaced line arrays (ULAs) and many other array designs. These beamformers include real-time array shape estimation (ASE), array shape correction (ASC), and range focusing. We have developed and implemented a number of adaptive beamformer (ABF) algorithms for Navy technology demonstrations and fleet systems.
The Robust Estimate and Plug (REAP) ABF algorithm was initially developed for the Advanced Deployable System (ADS) program and later used for a SURTASS TwinLine TB-29 demonstration test in conjunction with the Submarine Superiority Program. It is also being used in a torpedo defense application for small high-frequency volumetric arrays such as the ACI, AACI, and SPVA.
Another ABF algorithm developed at Adaptive Methods that has been applied in a wide range of Navy programs is the Short-Time Adaptive Broadband Beamformer, or STABB. This novel frequency-domain algorithm can adapt to the in-situ noise field in a single spectral update, making it suitable for adaptive spatial processing of highly dynamic signal and noise fields such as active pings in reverberation, and close passive engagement scenarios.
Adaptive Methods also developed a highly efficient beam-based ABF (TwinLine AIC) for the SURTASS TwinLine program that is currently integrated into the ARCI(I) system and in fleet use. We also integrated this algorithm into a beamformer we provided for the IUSS Common Processor (ICP) program for PMS-485. To recover lost performance due to dead sensors in uniformly spaced line-arrays (ULAs) but without resorting to the large increase in processor horsepower necessary with ABF, Adaptive Methods has developed a series of adaptive “hole-fill” techniques for PMS-485.
CBF
Adaptive Methods has developed a number of CBF algorithms with a variety of options that support a wide range of array designs. We are constantly improving beamforming algorithms to maintain a competitive edge. We have developed and implemented a frequency-wavenumber DeMuth beamformer for long ULA’s with highly efficient array shape correction (ASC), range focusing, and true-bearing stabilization options. In the DeMuth beamformer, ASC is implemented as a wavenumber domain convolution, saving significant processing power vs. a conventional dot product beamformer. We have also developed and implemented a class of conventional beamformer for arrays with arbitrary shapes, including options for length pruning, sub-aperture selection/formation, three-dimensional steering, pitch/roll compensation, and true-bearing stabilization. The AN/SQS-53C hull array beamformer Adaptive Methods has provided for A(V)15 Build-0 is a good example implementation of this type of beamformer. Lastly, as part of the ICP program for SURTASS mobile arrays, we implemented a length-pruned conventional beamformer with real-time array shape correction. This is followed by a beam-space adaptive beamformer, which is described in other sections.
Array Shape Estimation (ASE)
Adaptive Methods has been active in the area of ASE algorithm development for many years, beginning in the 1980’s, where we developed algorithms as well as compared a variety of different methods using beamformer output metrics. Later we developed ASE algorithms for SURTASS TwinLine (1993-1996), and Ardent (1999-2004), and have been active in developing and implementing ASE algorithms for the AN/SQQ-89 program. For TwinLine and Ardent, the shape is estimated from heading, depth, and acoustic Shape Measurement Units (SMUs). As an example, TwinLine shape is shown below using modeled data -- with and without errors. As the logical next step, we developed and implemented shape correction algorithms in several conventional beamformers, including SURTASS TwinLine. For details, see the CBF section.
Robust Estimate and Plug ABF (REAP)
In the late 1990’s Adaptive Methods (then Applied Hydro-Acoustics Research, Inc.) participated in an ABF algorithm development for the Advanced Deployable Systems (ADS) program sponsored by PMW-182. There was a competition among five organizations, all of whom brought an ABF to the table. After an 18-month competition, it came down to two algorithms with essentially the same performance, and one was REAP. Given the efficiency of the algorithm, REAP was chosen and implemented in real-time hardware for a four-array full-scale evaluation test (FET). This ABF uses an interactive white noise gain constraint to control main-lobe shape and sensitivity to mismatch. The implementation for ADS has the noise gain capability of a single gain constraint MVDR with the main-lobe control of a multi-point constraint algorithm, and only a 10% increase in processing load vs. a MVDR adaptive beamformer without white noise gain control. This algorithm, and variants, has been used for several advanced developments including a SURTASS TwinLine TB-29 demonstration acoustic trial and processing of a high-frequency acoustic intercept array for the AN/SQQ-89 program. The algorithm was implemented in real-time hardware with an exact hard-constraint option for a torpedo defense project. The figure shows how REAP controls the main-lobe and side-lobe response to loud contact for various white noise gain control settings.
Short-Time Adaptive Broadband Beamformer (STABB)
The Short-Time Adaptive Broadband Beamformer (STABB) was initially developed by Adaptive Methods for an active AN/SQS-53C hull array application under a Navy SBIR program, but STABB has now been successfully applied for AN/SQS-53C hull array passive beamforming in the IPS program, submarine broadband high-resolution noise source localization for NSWC-CD in the Source Localization and Analysis Work Station (SLAWS), high-resolution vertical localization in an active sonar application for mine-avoidance and also for a torpedo defense active beamforming application for ONR. The success of STABB comes from its ability to adapt rapidly. Implemented in the frequency-domain, STABB uses bandwidth (not time) for covariance estimation, allowing for extremely rapid adaptation. In addition, built-in rank reduction lets the algorithm be tailored for each application – trading degrees of freedom for execution speed. An integral robustness constraint is used to control the width of the main-beam, and the response to sources of mismatch. The ability of STABB to improve localization is dramatic. An example is shown in the figure below.
Failed Channel Recovery (CHRP-D)
Adaptive Methods has been focusing their efforts in developing algorithms to synthesize data for failed channels to recover array gain and side-lobe control. The most promising of these methods was a data-dependent channel repair method called CHRP-D. This algorithm adaptively estimates and fills dead channels using a linear least-squares estimator. The plot below shows a cumulative distribution of beam noise levels from a good ULA, a ULA with 10% of the channels dead, and ULA performance with CHRP-D hole-fill. Beam noise levels in quiet regions are filled in due to poor side-lobe control with dead channels, and this performance is recovered with CHRP-D. Adaptive Methods is currently implementing this algorithm in real-time hardware for inclusion in a future ICP build.
