The Information Systems Office (ISO) at DARPA develops, applies, integrates, and transitions information technology and systems to enable domination of the battlespace. To that end, ISO is engaged in three thrusts: comprehensive battlespace awareness; intelligent and timely force management and battle execution, and; realistic and affordable simulation for training, mission rehearsal, and course of action evaluation. In each thrust, ISO concentrates on enduring and future threats and solutions. The development approach involves creating the next generation of infrastructure, technology, and applications to build, sustain, and maintain a tightly-coupled system of systems. ISO information systems drive evolving concepts and doctrine for implementing a new warfare paradigm in which knowledge, not mass and fire power, is key to battlespace dominance across the ever-expanding spectrum of conflict. In all its operations, ISO programs foster cost effective acquisition of follow-on development and production systems. The following presents the goals of each ISO thrust: battlespace awareness; force management; and simulation.

Steel Rattler is a multi-phased project to determine the feasibility of using commercial off-the-shelf components in an advanced acoustic/seismic unattended ground sensor. This project is supported by the Defense Intelligence Agency through Sandia National Laboratories as the lead development agency. Steel Rattler uses advanced acoustic and seismic detection algorithms to categorize and identify various heavy vehicles down to the number of cylinders in the engine. This detection is accomplished with the capabilities of new, high-speed digital signal processors which analyze both acoustic and seismic data. The resulting analysis is compared against an onboard library of known vehicles and a statistical match is determined. An integrated thermal imager is also employed to capture digital thermal images for subsequent compression and transmission. Information acquired by Steel Rattler in the field is transmitted in small packets by a built-in low-power satellite communication system. The ground station receivers distribute the coded information to multiple analysis sites where the information is reassembled into coherent messages and images.

The unattended sensing of stationary (i.e. non-mobile) targets is important in applications ranging from counter- proliferation to law enforcement. With stationary targets, sources of seismic, acoustic, and electro-magnetic emissions can potentially be used to detect, identify, and locate the target. Stationary targets have considerably different sensing requirements than the traditional mobile-target unattended ground sensor applications. This paper presents the novel features and requirements of a system for sensing stationary targets. In particular, issues associated with long-listen time signal processing for signal detection, and array processing techniques for signal localization are presented. Example data and signal processing outputs from a stationary target will be used to illustrate these issues. The impact on sensor, electronic signal processing, battery subsystem, and communication requirements will also be discussed. The paper will conclude with a detailed comparison between mobile-target and stationary-target unattended ground sensor architectures.

This paper describes a passive unattended ground sensor which employs multiple sensor nodes of various transducer types to effect a sensor of exceptional utility and performance. This sensor will employ a seismic/acoustic classifying sensor and one or more additional sensor nodes as short or long range infrared, magnetic, piezo cable, microwave, or any other transducer providing a digital output. The resulting multi-node sensor will provide all weather capabilities to detect targets with high reliability, classify as either personnel, wheeled vehicle, or tracked vehicle, count the number of targets, and determine the direction of travel. The sensor weights less than two lb. and is one quarter the cost of comparable single-node-only sensors. The sensor communicates via a high quality RF link to a hand-held monitor which can be used alone, or tied to a palmtop/laptop computer to display detections on a digital map overlay. These capabilities allow the sensor deployment to be tailored by fielded forces to the terrain, environmental conditions, and threat to maximize mission effectiveness. The operational flexibility this provides a soldier is currently unmatched at any cost, at any size, and yet will be provided in a small, lightweight, low cost unit. The paper describes the design and performance of IDEWS-MN, the Communications Systems multi-mode unattended ground sensor system emphasizing the advantages of the modular architecture that allows for the inclusion of additional sensor capabilities, without changes to the basic hardware, software, or operational concepts of the sensor system.

This paper will review ongoing research in technologies that support Unattended Ground Sensors (UGS). Future requirements for UGS systems are offered based on observations of military evolution. We make the argument that UGS systems of the future will serve a broader functionality, including mobile soldier and small unit operation missions. Research results are then presented in the areas of acoustic classification, noise radar, low-cost IR imaging and low power techniques. The research we have undertaken shows promise for low-power, low cost sensors and signal processing in support of the UGS vision.

The U.S. Border Patrol monitors the traffic on the Mexican/U.S. Border, the Canadian/U.S. Border and along some coastal areas. Measures have been taken to reduce or eliminate illegal immigration and smuggling. An automated border surveillance sub-system based on the DARPA Internetted Unattended Ground Sensors Program is discussed.

Video sensors represent one facet of the unattended ground sensors family used to support military operations. This paper provides technical details on a modular, automated video surveillance (AVS) system concept that provides a 4D modeling and segmentation capability. This capability, combining 3D imaging and embedded information processing, enables the AVS to reliably detect and roughly classify objects and activities with a high probable certainty and minimal false alarms. The beneficial result is an elimination of the need for human monitoring and reduction in data transmission requirements by several orders of magnitude. This modular, forward observer AVS sensor package is comprised of integrated CMOS arrays and lenses (one for monoscopic view and two rigidly fixed for stereo), image processing, communication, GPS chips, and battery. This smart sensor is inexpensive, miniature, self-contained, and man-packable. Real-time video processing onboard the sensor provides: analysis and automatic target recognition algorithms enabling robust moving object detection; tracking and delimitation; and target characterization based on motion, size, form factor, texture, and specific identifying characteristics of objects. Automation of video screening tasks provides the benefits of visual surveillance without the associated burden or distraction from the mission. Because of its size and simplicity the AVS requires minimal man-in-the-field set up and lends itself to clandestine deployment.. In the future it may be possible to reliably and cost effectively air deploy the AVS. Each method offers great benefits as unmanned forward observers and sophisticated `tripwires' for battlefield awareness. While well suited for military reconnaissance, surveillance, and target acquisition missions for peacetime and/or wartime applications, the AVS also lends itself to other clandestine applications such as drug interdiction surveillance, monitoring, and tip off.

This paper describes the development of a frequency sensitive acoustic transducer that operates in the 10 Hz to 10 kHz regime. This device uses modern silicon microfabrication techniques to form mechanical times that resonate at specified frequencies. This high-sensitivity device is intended for low-power battery powered applications.

In the fall of 1995, a unique unattended ground sensor experiment was conducted at the Nevada Test Site. In the experiment, a variety of electro-mechanical equipment was operated, while data were gathered using a number of different types of unattended sensors at different locations. The sensors in this study included seismometers, accelerometers, electric dipole sensors, magnetometers and microphones. The purpose of this experiment was to gather data to explore and understand the performance of unattended ground sensor systems and the physical phenomena that can affect them. In this paper, we explore a few physical phenomena which can affect unattended ground sensor system performance.

Machinery typically generates mechanical vibrations at multiple, harmonically related frequencies which arise from various mechanically coupled moving components of machines or characteristic nonlinearities in their operational loads. These mechanical vibrations propagate from their origin through the air as acoustic waves and through the earth as various types of seismic waves. Of the two modes of propagation the seismic mode of propagation is the more complicated since the same harmonic may propagate simultaneously in various wave types (compressional waves, shear waves and various surface wave types) with differing propagation vehicles. Moreover, air-to-ground coupling has been shown to occur in some cases. The consequence of this multi-mode propagation is that standing wave interference patterns are set up over the terrain surrounding the sources which complicates the frequency-wavenumber analysis and identification of the signals. Since the set of harmonics omitted from a given type of machinery tend to be phase- coupled, higher order spectral analysis offers means for detecting and separating such coupled sets and reducing much of the Gaussian background noise and uncoupled sinusoidal noise components. In this paper we utilize sections through bispectral estimates obtained from continuous signals from various types of machinery with durations exceeding a minute.

This paper presents a technical approach for fusion of data from a distributed set of remote fusion nodes. Each remote fusion node contains a suite of unattended ground sensors (acoustic, magnetic, seismic, and environmental). A multiple-hypothesis fusion architecture is developed for fusing the output of the remote sensor outputs. Both target kinematic and target attribute data is processed for the purpose of localizing and classifying targets of interest. The higher-level fusion products of these remote fusion nodes (target state and classification) are then passed on to a central fusion node that also employs a multiple- hypothesis fusion architecture. With this hierarchical-based strategy, we illustrate how we obtain near-optimal fusion performance while maintaining a high degree of redundancy from potential loss of one or more of the remote fusion sites.

Advancements in both sensor hardware technology and in software systems and processing technology have enabled the development of practical realtime situation assessment capabilities based upon information from unattended ground sensors. A decision support workstation that employs rule- based expert system processing of reports from unattended ground sensors is described. The primary goal of this development activity is to produce a suite of software to track vehicles using data from unattended ground sensors. The situational assessment products from this system have stand-alone utility, but are also intended to provide cueing support for overhead sensors and supplementary feeds to all- source fusion centers. The conceptual framework, developmental architecture, and demonstration field tests of the system are described.

The event identification problem plays a large role in the application of unattended ground sensors to the monitoring of borders and checkpoints. The choice of features and methods for classifying features affects how accurately these classifications are made. Finding features which reliably distinguish events of interest may require measurements based on separate physical phenomena. Classification methods include neural net versus fuzzy logic approaches, and within the neural category, different architectures and transfer functions for reaching decisions. This study examines ways of optimizing feature sets and surveys common techniques for classifying feature vectors corresponding to physical events. We apply each technique to samples of existing data, and compare discrimination attributes. Specifically, we calculate the confusion matrices for each technique applied to each sample dataset, and reduce them statistically to scalar scores. In addition, we gauge how the accuracy of each method is degraded by reducing the feature vector length by one element. Finally, we gather rough estimates of the relative cpu performance of the forward prediction algorithms.

The infancy of unattended ground based sensors is quickly coming to an end with the arrival of on-board GPS, networking, and multiple sensing capabilities. Unfortunately, their use is only first-order at best: GPS assists with sensor report registration; networks push sensor reports back to the warfighter and forwards control information to the sensors; multispectral sensing is a preset, pre-deployment consideration; and the scalability of large sensor networks is questionable. Current architectures provide little synergy among or within the sensors either before or after deployment, and do not map well to the tactical user's organizational structures and constraints. A new distributed sensor architecture is defined which moves well beyond single sensor, single task architectures. Advantages include: (1) automatic mapping of tactical direction to multiple sensors' tasks; (2) decentralized, distributed management of sensor resources and tasks; (3) software reconfiguration of deployed sensors; (4) network scalability and flexibility to meet the constraints of tactical deployments, and traditional combat organizations and hierarchies; and (5) adaptability to new battlefield communication paradigms such as BADD (Battlefield Analysis and Data Dissemination). The architecture is supported in two areas: a recursive, structural definition of resource configuration and management via loose associations; and a hybridization of intelligent software agents with tele- programming capabilities. The distributed sensor architecture is examined within the context of air-deployed ground sensors with acoustic, communication direction finding, and infra-red capabilities. Advantages and disadvantages of the architecture are examined. Consideration is given to extended sensor life (up to 6 months), post-deployment sensor reconfiguration, limited on- board sensor resources (processor and memory), and bandwidth. It is shown that technical tasking of the sensor suite can be automatically accomplished via the warfighter's tactical direction enabling the DoD's vision of a `single logical taskable entity'.

Unattended ground sensors (UGS) utilize data from a variety of sensors (e.g., acoustic, seismic, and imagery) to make a determination about an unknown potential target. The Steel Rattler UGS derives its target identification solution from acoustic and seismic data. The identification solution and optional still imagery of the target are transmitted to the appropriate operating bases via satellite. This paper describes the various Steel Rattler hardware components used in the target identification process, the optional imaging system, and the communication system used for testing and demonstration purposes.

This paper examines a number of modulation, forward error- correction (FEC) coding, and synchronization techniques for use with unattended ground sensors. Included in the discussion are coherent and non-coherent modulation techniques, and block and convolutional codes. The synchronization requirements for coherent and non-coherent signaling are analyzed. it is shown that the estimation the signal phase for coherent demodulation does not increase length of the preamble used to aid synchronization. Hence, coherent modulation should be chosen over non-coherent signaling to improve the power efficiency. It is also shown that the use of FEC coding does not increase the required preamble energy. Thus, coding should be employed to achieve an improved bit error rate.

This paper describes the payload delivery system developed and proven to deploy an electronic warfare device to specific, predetermined locations on the battlefield. Initially called the Artillery Delivered Expendable Jammer (AD/EXJAM), it is now designated the Air Delivered-Ground (Deployed) Expendable Jammer (AD-G/EXJAM). The initial units were demonstrated from 155 MM artillery; the later units, from UAV's, helicopters and slow moving, fixed wing aircraft. While these two delivery systems were originally designed specifically for the EXJAM system, the concept is directly applicable to unattended ground sensors that require unmanned remote emplacement. Keys to the success of the jammer included design, development and field testing of power supplies, antennas, deployment systems and packaging to allow payloads to withstand high-g impact and other severe environments typically encountered. The artillery deployed systems were designed to be `wooden' rounds needing no special handling and storing. These systems treat the payload as independent elements which are self-ejected from a fired M483A1 or M864 round and are completely automatic upon hitting the ground. The more recent payloads can be delivered from UAV's and include remote control capabilities, increased operating life and increased power output. The present payload is packaged into a cylindrical shape, approximately six inches in diameter and 6.5 inches long and are contained within a carrier, attached to an Unmanned Air Vehicle (UAV) or any other air vehicle. Upon reaching the dispensing point, the release command can be issue by either the UAV or a separate ground control unit in RF contact with the carrier. The carrier then begins a timed dispensing sequence that has been selected for optimum payload emplacement in the target area. New developments include a design and subsystem demonstration of a tactical munitions dispenser variant of the deployment system. Operational characteristics of any specific sensors or payloads will not be addressed.

The effectiveness of sensors to identify, locate and characterize a facility depends on several factors, namely, the types of equipment being monitored, the location of the equipment, the mounting of the equipment to the structure, the physical configuration of the facility, the surrounding propagating media, and the effects of natural and cultural background noise. We have developed electromagnetic, seismic, and acoustic source and propagation models to theoretically investigate these issues for facility models in realistic geological media. Source models for various machine types have been developed along with an electromechanical system parameter database. The model produces expected time-varying currents on the power lines which generate electromagnetic fields observed by EM sensors. The time-varying torque function defined in the model gives the expected force time function for the seismic and acoustic source. Propagation models have been developed, using 2D seismic, acoustic and electromagnetic finite- difference techniques, for propagating the fields through the structure and earth medium to far-field sensors. In this research, we are also investigating methods for fusing EM and seismic/acoustic signatures for more complete characterization of the operational elements in a facility. The products of the effort are enhanced key components of a forward modeling system incorporating a CAD-style interface for specifying facility models, equipment locations and surrounding media. We envision the system as a tool to estimate detection ranges and optimal sensor placement for monitoring a facility.

Sandia National Laboratories has recently developed two major field test capabilities for unattended ground sensor systems at the Department of Energy's Nevada Test Site (NTS). The first capability utilizes the NTS large area, varied terrain, and intrasite communications systems for testing sensors for detecting and tracking vehicular traffic. Sensor and ground truth data can be collected at either of two secure control centers. This system also includes an automated ground truth capability that consists of differential Global Positioning Satellite receivers on test vehicles and live TV coverage of critical road sections. Finally there is a high-speed, secure computer network link between the control centers and the Air Force's Theater Air Command and Control Simulation Facility in Albuquerque NM. The second capability is Bunker 2-300. It is a facility for evaluating advanced sensor systems for monitoring activities in underground cut-and-cover facilities. The main part of the facility consists of an underground bunker with three large rooms for operating various types of equipment. This equipment includes simulated chemical production machinery and controlled seismic and acoustic signal sources. There has been a thorough geologic and electromagnetic characterization of the region around the bunker. Since the facility is in a remote location, it is well-isolated from seismic, acoustic, and electromagnetic interference.

In many applications, sensors can be precisely placed to optimally achieve their intended monitoring function. However circumstances arise in which sensors must be delivered to a remote location by air drop or from gun launched projectiles due to presence of hostile forces and environments or unavailability of resources for their accurate placement. Consequently, sensors are distributed in a random pattern. This paper presents an approach to locating each sensor in the net from a remote location. It further describes an approach to propagating the coordinates to a remote control point and identifying the coordinates of sensors activated by intrusion.

In times of war, precision targeting of enemy forces requires timely updates of their movements. Overhead assets may be impaired by natural and human origin interference. Internened Unattended Ground Sensors will augmentthe overhead assets and ensure constant, accurate surveillance updates to allied commanders. The emplacement of Internetted Unattended Ground Sensors (TUGS) sensors by Tomahawk deep into enemy territory will allow the destniction of threat forces early in conflict reducing friendly casualties and reducing the enemy's offensive capabilities. Deep delivery of JUGS by the Tomahawk platform will reduce risk to military personneL lower cost, and ensure accuzate delivery to watchdog locations

In this paper, we highlight several robotic and sensor related technologies that will enhance military operations in the 21st century. Aspects of these technologies will have mission impact in both urban environments and rural settings. We describe technologies to aid humans in intelligence, threat warning, surveillance, reconnaissance, and target strike operations.

Confidence building between regional parties can be facilitated through cooperative seismological research activities. Shared data, facilities, technology, and research results can (1) assure participants that nuclear testing is not taking place, (2) provide information that can be used to characterize the geophysical parameters of a region for earthquake hazard mitigation, and (3) support basic seismic research.

The goal of the test bed project is to test portable, unattended chemical analysis instruments that will help verify the compliance of various international agreements on weapons of mass destruction. We report on the design of the test bed and present response curves for the first sensor deployed at the test bed. The architecture of the data- acquisition and display interface utilizes industry standards (LonWorks and CORBA), state-of-the-art developmental tools, advanced data visualization and display tools, and commercial government off-the-shelf software and hardware in order to have a flexible/modular infrastructure for integrating and testing both sensors and software applications for unattended remote monitoring systems. The HAZMAT Spill Center located at the Nevada Test Site will be described as well as the opportunities it offers for testing unattended chemical sensors.

Sensor arrays offer opportunities to beamform, and time- frequency analyses offer additional insights to the wavefield data. Data collected while monitoring three different sources with unattended ground sensors in a 16- element, small-aperture (approximately 5 meters) geophone array are used as examples of model-based seismic signal processing on actual geophone array data. The three sources monitored were: (Source 01). A frequency-modulated chirp of an electromechanical shaker mounted on a floor of an underground bunker. Three 60-second time-windows corresponding to (a) 50 Hz to 55 Hz sweep, (b) 60 Hz to 70 Hz sweep, and (c) 80 Hz to 90 Hz sweep. (Source 02). A single transient impact of a hammer striking the floor of the bunker. Twenty seconds of data (with the transient event approximately mid-point in the time window). (Source 11). The transient event of a diesel generator turning on, including a few seconds before the `turn-on time' and a few seconds after the generator reaches `steady-state conditions'. The high-frequency seismic array was positioned at the surface of the ground at a distance of 150 meters (North) of the underground bunker. Four Y-shaped subarrays (each with 2-meter apertures) in a Y-shaped pattern (with a 6-meter aperture) using a total of 163-component, high- frequency geophones were deployed. These 48 channels of seismic data were recorded at 6000 and 12000 samples per second on 16-bit data loggers. Representative examples of the data and analyses illustrate the results of this experiment.

The tactical motivation for a compact and easily deployable remote sensing unit for hazardous material releases is described. The most important measurements such a unit could make for accurately determining the extent of a plume from a target or potential target is a range resolved profile of the wind within the vicinity of the strike zone. However, to make such measurements a new type of instrument is required which is inherently robust, compact, and efficient. To demonstrate that such an instrument is feasible, we have designed, built, tested, and performed an initial laboratory evaluation of a compact Doppler lidar system for tactical applications.

Commercially available Digital Signal Processors can be used to host state-of-the-art air acoustic adaptive beamforming algorithms in low power, low cost, real-time sensor systems. These systems are suitable for use as both unattended ground sensors and in platform-mounted applications. This paper describes a compact state-of-the-art, real-time adaptive beamforming approach and sensor hardware. Recent day/night field test results for detection range, multiple target tracking, and classification are presented for various vehicles. The data focuses on long range target detection as well as tracking and classification performance in multiple target environments composed of closely spaced or clustered targets. Target location (x-y position) performance using real-time netted sensors (sensor fusion) is also presented.

The U.S. Army Research Laboratory (ARL) is developing an acoustic target classifier using a backpropagation neural network algorithm. Various techniques for extracting features have been evaluated to improve the confidence level and probability of correct identification. Some techniques used in the past include simple power spectral estimates (PSEs), split-window peak-picking, harmonic line association (HLA), principal component analysis, wavelet packet analysis, and others. In addition, improved results have been obtained when data are combined from other sensors co- located with the acoustic sensor. A three-axis seismic sensor has been configured as part of an acoustic sensor array that ARL uses on typical field experiments, with data collected and sampled simultaneously. The PSE, HLA, and shape statistic feature data are extracted from a group of vehicles and then split into testing and training files. The training file typically consists of 75 percent of the data set, and the performance of the trained neural network is evaluated with the remaining test data. Cross-validation is performed with vehicle data collected at different times of delay and under various conditions. Results of the neural network from a few of the feature extraction algorithms under evaluation and from the fusion of the acoustic and seismic sensor data are presented.