Technology for the United States Navy and Marine Corps, 2000-2035: Becoming a 21st-Century Force: Volume 2: Technology (1997)

In the future, affordable automated systems and equipment will supplement or supplant many current operator functions in routine operations both aboard ships and on the shore. Automation technologies will also be employed in many steps in the chain from target search and acquisition to weapon impact in an effort to reduce platform size, manpower, and budgets and to minimize the number of personnel exposed to hostile action. Key areas involving automation technologies include (1) unmanned underwater vehicles (UUVs), (2) unmanned aerial vehicles (UAVs), (3) ship automation, (4) platforms and weapons guided by integrated inertial navigation systems (INSs) and the global positioning system (GPS), (5) robotics, and (6) automatic target recognition (ATR).

Opportunities for automating operator functions range from mundane housekeeping and monitoring tasks to intelligent robots for performing hazardous or superhuman tasks. In the time frame of reference, the panel believes that technology will permit semiautonomous, remotely operated combatants as well as fully autonomous low-observable combatants with a variety of tactical capabilities, themselves controlling smart UUV and UAV off-board sensor networks for an extended spatial and temporal awareness of the surrounding space. These unmanned, independent platforms will benefit from all-weather continuous communications and autonomous sea-keeping and logistics support.

The reduction of operators will allow naval and expeditionary platforms to become significantly smaller, more robust, and survivable for the same sensor and weapons load. Many current operators will be replaced with internal smart sensors teamed with robotic devices and highly reliable, fault-tolerant hardware

and software systems to deal with casualty and damage detection, control, correction, and recovery.

UUVs and UAVs will go beyond just serving as off-board sensor platforms. Some will be capable of long-term independent operations. Others will combine sophisticated target detection, classification, and tracking capability with maneuvering and weapons-launch capability to serve as semiautonomous and possibly autonomous strike or fleet defense platforms. Those serving as off-board sensors will possess all-weather autonomous launch-and-recovery capability, with the mother platform able to provide automated logistics support.

Platforms, UUVs and UAVs, and weapons will enjoy the benefits of continuously available position information to 1 m or less accuracy from small, inexpensive integrated GPS/INS navigators. Sophisticated adaptive mission planning/replanning processors and ATR software on board these semiautonomous vehicles and weapons will support obstacle avoidance and mission restructuring in real time.

Unmanned underwater vehicle technology is actually a combination of a number of emerging technologies that can provide the naval forces with new operational capabilities in undersea warfare, reconnaissance, and surveillance. During the latter days of the Cold War, UUVs were envisioned as force multipliers for the blue-water antisubmarine warfare (ASW) mission, including mobile acoustic barrier patrol, off-board active acoustic sources, deep-water mine avoidance for submarines, and tactical aids for port egress. Although considerable research was undertaken and a number of operational demonstrations were conducted in the late 1980s and early 1990s, the blue-water ASW problem was perceived to disappear before any operational UUVs were deployed. In the post- Cold War era, potential UUV missions have centered on undersea littoral warfare, including shallow-water mine reconnaissance, diesel submarine ASW, surveillance, and counterproliferation. The Department of the Navy has just started an acquisition program for the first operational UUV, the long-term mine reconnaissance system (LMRS), which is a submarine-launched and submarine-recovered countermine system.

UUVs can conduct a series of primary and support missions to extend the fleet's warfighting capability, provide surveillance and reconnaissance functions, and conduct tactical oceanographic missions. UUVs not only provide increased mission capability but also should not add significantly to operational costs, since

no crew is required and, for the most part, maintenance and support can be conducted by existing ship's force.

There is no current tactical application of UUVs. The various technologies are just becoming mature and operational confidence is growing. The Department of the Navy has just started its first UUV acquisition program, LMRS.

UUVs will be operated by both submarines and surface combatants, and variants may well be launched from fixed-wing and rotary aircraft. Autonomous, independent UUV operations, without a host platform, are also likely for certain missions. UUVs will provide economical force multiplication, increased battle-space awareness, reduction in exposure to hostile action, and extend operational capability.

Future UUV applications will include the following:

• Off-board sensing. Platforms for acoustic and nonacoustic detection and localization and tracking of targets. With multiple vehicles operating as nodes in a far-ranging, cooperative, smart-sensor network, detection thresholds will be significantly improved.
• Intelligence collection. Covert platform for gathering tactical and strategic intelligence, including minefield surveillance, harbor and waterway traffic surveillance, surveys of uncharted bodies of water, plus the ability to carry a limited range of weapons for attacking detected targets.
• Mission support. Autonomous, covert, and stealthy electronic warfare (EW) and information warfare (IW) platforms for decoy operations in coastal areas, for example.

The most significant UUV technology needed for long-duration (30 days or longer) autonomous operation is safe, reliable, high-energy-density and high-power-density batteries and fuel cells operating from seawater or equivalent sources of energy. Of almost equal importance is the need for compact, cost-effective, reliable UUV hardware and software systems.

On-board intelligent navigation and control and decision planning and re-planning are essential technologies for reliable, long-term unattended operations. Other important technologies include low-power, high-performance acoustic and nonacoustic sensors; high-bandwidth, continuous, convert communications between cooperating UUVs, between UUVs and hosts, and between UUVs and

global-reaching operators; and automated resupply, launch, and recovery from unmanned hosts.

Low observability to avoid acoustic and nonacoustic detection is essential to the survivability of autonomous UUVs. This requires vehicle shaping, acoustic and optically absorbing material coatings, equipment quieting, and low-noise propulsion, such as direct electric drive.

Current commercial applications focus more on remotely operated vehicles (ROVs) rather than on autonomous vehicles. Because these vehicles share some of the same needs as the autonomous UUVs envisioned for naval force operations, a small portion of the technical advances in the above technologies can be expected from the commercial sector. Technologies for long-duration, highly reliable autonomous surveillance and tactical missions will need to be developed by the Department of the Navy.

In general, foreign developments in UUV technology have lagged significantly behind the work conducted in the United States. The United States has made a significant technology investment in the military application of unmanned undersea vehicles, whereas most foreign UUV research has been at the university level, and much has been oriented toward scientific and oceanographic uses of the technology. The only exception is the former Soviet Union, which did invest in UUV research and development for the military.

The first operational UUV, the LMRS, is scheduled for initial operation in 2003 to 2004. It is likely that other UUV systems would be easily inserted within the same time frame. More sophisticated systems with higher degrees of autonomy enabled by intelligent software systems that can plan and replan to deal with failures and contingencies will follow in subsequent years. Toward the 2035 time frame, the panel can envision highly autonomous UUVs that operate in cooperative engagements. These societies of UUVs will be capable of sensing their environments and communicating with each other to optimize underwater missions.

Unmanned aerial vehicles are small to large unmanned aircraft that can be launched from ships and shore to conduct reconnaissance and surveillance, to coordinate and participate in cooperative engagements against targets, and to act as weapons in some applications.

UAVs have the potential for becoming pervasive solutions for many future Navy and Marine Corps needs. The rapid evolution of mission-payload technologies, including sensors and sensors combined with weapon systems, and aeronautical technologies, including navigation, autonomous control, and propulsion, will enable UAV missions heretofore not possible.

A future vision of the UAV for naval operations might be as follows: Navy and Marine Corps forces will control a host of UAVs sweeping the combat areas around the battle group and onshore with electro-optical/infrared (EO/IR) and synthetic aperture radar (SAR) reconnaissance sensors, collecting enemy radar and communications signals, coordinating friendly communications, detecting mine fields, enemy submarines, and cruise missiles, jamming enemy radars and communications, and performing targeting and control missions for precision weapons.

Linking several UAVs with precision timing into a society of agents would create a signals-collecting dragnet, not allowing a single RF signal to escape collection and geolocation. Similarly, UAVs of the future will be able to deliver the appropriate response, whether it is jamming, performing as an antiradiation weapon, generating directed-energy weapon (DEW) disruption signals for information warfare purposes, or delivering precision-guided munitions.

Although most UAV systems envisioned today involve remote piloting and require some takeoff-and-recovery real estate, future UAVs can well use short takeoff and landing (STOL) or vertical takeoff and landing (VTOL). They will be totally autonomous with sufficient on-board intelligence to change their operational behavior on the basis of observations in the conflict areas. Strategic intelligence will be acquired with high-altitude, long-endurance (many days), low-observable UAVs. They will be refueled in flight by tankers or perhaps by beams of microwave energy from ships or shore.

Most current UAV technology development is being carried out by the Joint DOD Cruise Missile and UAV program office, administered by the Department of the Navy. This office was established at the request of Congress to ensure that UAV development programs were coordinated and nonduplicative between military departments. In addition, DARPA is funding the development of high-altitude, long-endurance UAVs.

The panel believes that the highest priority should be given to the development of gas turbine engines for propelling UAVs. Gas turbine engines have long life and low weight-to-thrust ratios, use heavy fuels, and possess low acoustic signatures.

High-altitude and long-endurance operation would enable UAVs to continually observe the entire sphere of importance around the battle group and the expeditionary forces on shore. Payloads equipped with sensors sensitive over the spectral range from EO/IR to millimeter waves with radar and SAR and with electronic surveillance devices would observe all enemy activities and report this intelligence immediately to all friendly users. These UAVs would participate in cooperative engagements against difficult targets, such as low-observable missiles. The ability to time-integrate weak ocean-surface signatures from ships, low-flying missiles, and submarines would greatly increase the detection of these threats. With the future availability of affordable and reliable solid-state blue-green lasers, direct detection of mines and enemy submarines in the littorals and direct communications with friendly attack submarines and UUVs will become possible.

Because of the importance of these UAVs to surveillance and detection of hostile threats and targets, the panel strongly encourages the continual support of technology development in high-altitude, long-duration flight and in low-cost, small-size sensor payloads. Low-cost communications between UAVs and the user community will also need to be developed.

An interesting new airframe technology that should be considered for UAVs is the so-called free-wing. This wing configuration is free to rotate in pitch, providing for a completely passive means to adapt to the instantaneous direction of air flow. Thus, as the wing encounters turbulence, the attitude reconfigures without the necessity of active control. Not only does this technology allow for much higher platform stability because the entire airframe is not buffeted by turbulence as with a fixed wing, but also the wing is by its design stall-free, thus mitigating one of the major sources of aircraft failure.

The free-wing combined with another new concept called the tilt body, which adds trim tabs to the fuselage, provides a capability for small-area launch and recovery as well as a simple transition from helicopter-like flight to conventional flight. UAVs with VTOL capability would bring a new dimension to Navy and Marine Corps missions. The basic free-wing concept has been flight demonstrated both as a UAV airframe and as a manned ultralight vehicle. The free-wing tilt body has been built and flown successfully but has not yet been commercially or militarily applied. The use of this technology could lead to UAVs that can operate more effectively from a variety of naval ships. Both analytical and hardware designs are required, prototypes need to be built, and in-fleet experience needs to be gained.

UAVs with very low observability (VLO) features will be required for a variety of covert missions, and all UAVs must possess some degree of low observability for survivability. The panel believes that attention is needed to developing the low-cost materials and active signature management technologies that will enable these capabilities.

Efforts should also be undertaken to develop a family of small UAVs that are

significantly less expensive, so inexpensive that they could be considered as expendable. Such UAVs would open up entire new operations. Equipped with the low-cost, smart sensor-on-a-chip systems, as described in Chapter 4, and modest munitions, these micro-UAVs could perform functions such as search and find, followed by homing and kill.

The surface combatants, submarines, and amphibious and auxiliary ships of the future naval forces will have a much different configuration than the ships of today. Ships will be highly automated and conduct extensive semiautonomous (limited need for remote operator intervention or decision) operations. Many of the functions currently conducted by personnel on board today's ships will be performed by autonomous systems on board or in some cases teleoperated by humans via communication links. The reduced or nonexistent manning that results from the broad application of automation will allow smaller, more damage-resistant and survivable ships, with fewer personnel being exposed to hostile action.

The set of technologies required for full automation of a naval ship is diverse in scope and includes the following:

• Over-the-horizon (OTH) communications that are jam resistant, damage tolerant, and all-weather capable;
• Distributed computer systems with high-performance capabilities and architectures that are fault and damage tolerant;
• Advanced computer algorithms and programs for operation, monitoring, control, maintenance, and logistics of all combat, weapon, hull, mechanical, and electrical systems;
• Robotics for damage control and special functions such as under way replenishment and launch and recovery of UAVs and UUVs;
• Reconfigurable power-generation and auxiliary systems that are highly reliable, fault tolerant, and damage tolerant;
• Navigation technologies that are continuously all-weather capable and jam resistant and provide passive knowledge of the geographic location of a ship to 1-m accuracy; and
• Security technologies for physical, communications, and control systems.

The implementation of ship automation will significantly reduce manning, currently the highest cost element of naval ship operations. The reduction of personnel aboard ships will also reduce casualties in future hostile actions.

The fiscal reality of the high costs associated with people on ships has forced efforts within the Department of the Navy to reduce manning. The Navy's new surface combatant has a goal of 95 people compared with a manning level of about 320 on an equivalent ship today. The arsenal ship has established a goal of 50 people maximum. The Navy's next aircraft carrier will attempt to reduce the manning to one-half the present manning levels. The smart-ship initiative and some advanced technology demonstrations will make some progress toward removing people from ships, but much further work is needed. Significant reduction of people from ships requires not only definition and development of the technologies required but also acceptance of the manning philosophy and doctrine to make autonomous ship operations a reality.

The technology status and the developments necessary for autonomous ship operations are as follows:

• OTH communication capability exists today; however, the issues of being noninterruptable, jam resistant, damage tolerant, and all-weather capable must be addressed. The criticality and reliability of the communication link are overall system issues that will determine the extent and reliability of autonomy.
• The computer-processing power required to enable ship automation is expected to continually increase over the next 40 years and is not envisioned to be a limitation. The development of computer architectures and systems that are fault tolerant and damage tolerant is required, however.
• Advance computer programs and algorithms will require a large amount of development work. The philosophy and methods developed for the control of small autonomous systems, such as present-day UAVs and UUVs, will have to be greatly expanded and enhanced to address all the complex issues required for a ship to carry out extremely complex and lengthy missions.
• The rudimentary diagnostics used today to implement condition-based maintenance will have to be further developed into sophisticated prognostics in the future. Such prognostics must deal not only with the condition of machinery and equipment but also with the condition of the overall system, the impact of this condition on the warfighting capability of the ship, and the ability to reason, reach decisions, and take appropriate action.
• Robotics systems for damage control will require some development. Some of the technology development utilized for such things as factory automation may be adaptable to many of the damage-control and special applications on a ship.
• The ability for a ship to navigate anyplace on the globe in a nearly hands-off mode exists today. Expected advancements in antijam capability in GPS

• receivers and integration with inexpensive inertial navigation systems will permit the ship, the off-board sensors, and the weapon systems to precisely locate themselves at all times and under all circumstances. Automated mission planning and on-the-fly replanning using the certain knowledge of current geographic location will enable autonomous obstacle avoidance and mission reconfiguration in the future.

The Precise Positioning Service (PPS) of the Global Positioning Satellite system currently provides spherical error of position (SEP) accuracy to 16 m and CEP accuracy between 8 and 10 m over most parts of Earth. This impressive technology enables exceptional worldwide navigation, especially when multiple GPS measurements are combined in a Kalman filter to update an inertial navigation system on a platform or a weapon. The Kalman filter provides an opportunity to calibrate some of the GPS errors, such as clock and ephemeris errors, as well as several of the inertial system errors, and when properly implemented, CEPs of better than 8 m have been observed.

The ability to provide precise navigation for Navy and Marine Corps platforms anywhere in the world and to deliver precision weapons on target through combinations of external GPS and internal INS navigation is essential to future naval operations.

The path to the projected 1-m accuracy in the 2010 time frame includes the use of additional GPS ground monitor stations and advanced ground-based software to generate and uplink more accurate and frequent corrections to satellite clock and ephemeris data. Interranging and communications between satellites will be used to further refine the navigation information. Platform receivers will also use all-in-view tracking and will be navigating using 8 satellites and perhaps up to 12, as opposed to the 4 satellites in many current implementations. Advanced atmospheric models and algorithms will also be implemented to compensate for tropospheric and ionospheric effects, as well as multipath signals to the receiver.

In military applications, users are also concerned with the integrity and reliability of the system in the presence of spoofing, jamming, and interference. In the future, communication crosslinks between satellites and improved ground

monitoring of satellite health status will allow almost instantaneous information to users about failing satellites whose signals should be ignored. The widespread use of all-in-view receivers will allow continued high-accuracy navigation even with fewer satellites. In addition, the tight-coupling between the inertial navigation systems and the GPS data in future platforms will allow continuous inertial navigation with tracking from just a single satellite. This capability is particularly important for the Marine Corps ground forces and in low-level aerial applications. Low-level flight creates severe signal masking because of foliage, terrain, and urban structures. The continued use of encryption and the ability to rapidly change the military code will minimize false satellite navigation signals.

Jamming and intentional interference will remain serious issues for military as well as civilian operations. The GPS satellites are typically in high-altitude 12-hour orbits and broadcast at low power (on the order of 20 W). Thus, very low power jammers on or near Earth's surface can jam the weak received signal. Inertial navigation systems, however, are not jammable, and so there is real synergy in combining satellite and inertial navigation. Through the use of inertial system information, the tracking loop bandwidth in the GPS reception can be narrowed by an order of magnitude, thus increasing the antijam margin by 10 dB. In designing an integrated GPS/INS system, tradeoffs must be made between the accuracy and cost of the inertial system and the cost associated with an antijam GPS receiver/antenna capability.

Proponents of inertial systems argue that a high-antijam GPS receiver is not required because of the inherent high accuracy of the inertial information, whereas GPS proponents argue that use of a high-antijam GPS receiver will substantially reduce inertial system accuracy requirements. Both arguments are strongly dependent on the usually poorly defined mission and jamming scenarios. However, what has generally become accepted is that GPS is remarkably vulnerable to jamming during the signal acquisition phase with conventional receivers. For example, a broadband CW jammer with 1-W effective radiated power located at 100-km line-of-sight distance from a GPS receiver antenna would prevent acquisition of any satellites using the civilian coarse/acquisition (C/A) code. A 1-W jammer is cheap and is about the size of a hockey puck. Furthermore, the current civilian code can be spoofed by an even smaller power jammer. In general, a GPS receiver cannot be expected to acquire the C/A signal in a hostile environment. For long-range navigation applications, the C/A signal acquisition could be accomplished outside hostile territory, with a transition inside hostile territory to military P(Y) code lock that has a higher level of jamming immunity. A 1,000-W jammer at about 100 km would now be required to break receiver lock. If the vehicle approaches within 10 km of the jammer, power levels of about 10 W would be effective in breaking military code lock.

The military P(Y) code (an encrypted precision code) has more antijam protection than the civilian code because of its 10-times-larger spread-spectrum bandwidth. Therefore, it is important to develop receivers that can acquire the

military code without having to first acquire the civilian code. Because the military code is very long, however, many seconds of time and many correlators are needed for a two-dimensional search over code timing and Doppler frequency. It would be faster if satellite ephemerides and accurate code timing were available to perform a ''hot" start. For a GPS-aided weapon launched from an aircraft, accurate timing and satellite position could be transferred from the aircraft to the weapon. This transfer normally requires a wideband data bus, but few aircraft are currently so equipped. However, in the future, most aircraft will be so equipped and weapons will normally be hot-started directly into the military code.

As new receiver technology with massively parallel correlators, improved algorithms, and adaptive or nulling antenna technologies are incorporated into weapon and platform systems, antijam performance will improve significantly. If antijam performance is doubled (in decibels), then the jammer in the previous cases would have to be five orders of magnitude more powerful to be effective. Such powerful jammers would present inviting targets to antiradiation homing missiles. In the terminal flight area, the GPS receiver will probably be jammed and the weapon or platform will have to depend on inertial-only guidance or the use of a target sensor. Thus, it is important to make sure that adequate backup vehicle guidance and navigation capability is provided to meet military mission requirements against adversaries who are willing to invest in electronic countermeasures. This is true today and is expected to remain so in the foreseeable future.

To prevent the use of GPS by hostile forces, it may be necessary to jam the C/A code in the battlefield area, thus giving more emphasis for our own forces to directly acquire the P(Y) code. Additionally, a hostile platform using a combination of correction signals transmitted from a fixed ground station and the direct GPS signal to improve accuracy, a technique known as differential GPS, is vulnerable to jamming of the data links and destruction of the ground station. Many of these notions have been described in the 1995 report by the DSB Task Force on GPS.¹ The Russian GLONASS system currently provides accuracy similar to GPS-PPS, and hence some requirements could also be satisfied by using GLONASS under jamming conditions.

In the next few decades, accuracy of the integrated navigation solution will improve from the current 8-m to the 3-m range (for precision strike applications) and then to the 1-m range. Planned programs by the GPS Joint Program Office

(JPO) are in place to accomplish the latter. Complementary techniques to provide geo-location targeting information at comparable accuracy to the GPS/INS solution will be developed. Techniques to deny the use of accurate GPS (or the Russian equivalent GLONASS system) to hostile forces will also be developed in the next decade. In parallel, lower-cost inertial components with improved accuracy will be developed using MEMS technology and other approaches. Tightly integrated architectures for high-antijam GPS/INS systems will become common, replacing avionics architectures based on functional black boxes where receivers and inertial systems are treated as stand-alone systems. Techniques will be developed to deny the use of GPS or GLONASS by hostile forces.

A robot can be defined as a general-purpose machine system that, like a human, can perform a variety of different tasks under conditions that may not be known a priori. Robots are intelligent, taskable mechanical entities that can assist or replace sailors and marines performing high-risk operations in hazardous environments. Over the last decade, robotics technology has made significant inroads into high-volume manufacturing operations. Recent improvements in sensing, artificial intelligence (AI), and control technology, coupled with a significant improvement in low-cost, low-power computational devices, have made possible intelligent robots capable of operating in unstructured environments.

The key elements of robot systems are (1) effectors, which corresponds to the arms, hands, legs, and feet of humans; (2) sensors, which detect the conditions around the robot; (3) computers, which bring intelligence to the performance of the robot tasks; and (4) communications systems, which allow robots to communicate with other robots and with humans.

Robot systems are deployed and operated in different ways depending on the application as described below:

1. Teleoperated systems are devices that are controlled directly by a human operator. Such systems are also called remotely operated. The operator is usually connected to the device over dedicated or shared telecommunication links that allow control signals to be sent to the robot and allow sensor signals from the robot to be displayed to the operator.
2. Telepresent systems are specialized teleoperated systems in which the operator's perception of the remote environment is identical to what it might have been if the operator were present at the robot site. Achieving telepresence requires the use of immersive displays, high-resolution sensors, and much higher communication bandwidths.
3. Telerobotic systems involve the transfer of human control to autonomous

3. robot control sometime during the course of activities. For example, the operator may position the robot at the start of a task and then turn over control to perform the task to the robot.
4. Robot systems are the pinnacle of robot complexity. These systems can perform tasks autonomously under task guidance by human operators.

Robotic systems have been used extensively in repetitive manufacturing environments where their speed and precision offer significant advantages over human performance. Autonomous and teleoperated vehicles (ground vehicles, underwater vehicles, and aerial vehicles) allow extension of human presence into environments that are dangerous, e.g., battlefields, or inaccessible by other means, e.g., deep oceans.

Robot systems manipulate and navigate through the environment using actuators. Most robot manipulators are built as a collection of articulated links with a firm stable base and ending in a specialized end effector. End effectors can be specialized tools, such as a welding torch, or generic articulated grasping devices. Most ground-based mobile robots use tracked or wheeled platforms, although legged devices have significant advantages in harsh terrain, e.g., on beaches or in urban debris-filled buildings.

Until recently, design efforts were targeted at building large actuators with very high power-to-weight ratios, resulting in heating and vibration problems encountered during operation. Recently, however, there has been a shift toward building small, cheap, dispensable systems. Current actuation technology, however, is far from what might be required for the realization of small machines. This has resulted in a significant interest in the miniaturization of electric motors, as well as alternate means of actuation that might be applicable to small systems. The latter has resulted in the development of materials that might serve as artificial muscles, expanding gels, and electrostrictive polymers. And finally, there is a growing body of results related to microactuation, and a small number of actuator technologies have actually been scaled down to the micrometer scale. These technologies have been applied to a small number of problems, but the true challenge will be to integrate them into a centimeter-scale system using MEMS and other approaches to demonstrate coordinated micro/macro actuation.

Sensing is an integral part of most machines. It is used for taking the measurements required for controlling the machine. Currently, in addition to

various position and velocity sensors used for providing internal state information, the use of IR or ultrasound sensors for the detection of obstacles is quite common. Sophisticated systems also use one or more visual sensors for acquiring information about the nature of a robot's environment. And finally, a wide array of domain-specific sensors, such as subsurface imaging sensors or magnetic sensors for mine detection, or thermal-imaging sensors for detection and classification of humans, have been suggested for successfully tasking the robots within various scenarios.

Recent advances in stereo-vision algorithms coupled by low-cost, high-speed digital signal-processing chips have resulted in the ability to build extremely cheap vision sensors that can provide future robots with a very-high-resolution sense of its environment.

Control typically refers to the choice of actuation signals that might be required for forcing a machine to exhibit some prescribed behavior. Although traditional control has extensively studied problems related to set-point and trajectory control of linear systems, few formal concepts exist for the synthesis of control methods that have the ability to integrate information from a range of sensors, such as vision and touch. This has resulted in the investigation of alternate methods such as intelligent control, fuzzy control, neural-network control, and rule-based control for autonomous systems. Intelligent control refers to the ability of a control system to make decisions on line and to reconfigure control structures as necessary. Rule-based control attempts to express decisions as heuristically derived rules driven by sensory information. Fuzzy control applies fuzzy membership functions for the determination of control structures, and neural-network control uses neural networks for functional approximations in the identification and control processes.

With the advances in the enabling technologies outlined above, current robotics research is focused on the development of taskable, cooperative, autonomous devices in a variety of environments. The Department of the Navy is funding the development of a variety of semiautonomous underwater vehicles. In collaboration with the Army Unmanned Ground Vehicle office, robot vehicles are being investigated.

In the future, collaborative systems consisting of specialized robots at multiple scales, for example, miniature robots controlled by a larger mother robot, communicating with each other and with external controllers in very complex environments, will be needed in many military scenarios being discussed today. Research into computationally tractable representations of the environment, reasoning with uncertain and partial knowledge, cooperative tasking of heterogeneous devices, techniques for manipulation and locomotion in hazardous terrain,

sensing systems, robot power systems, and system-integration techniques will be required to bring these technologies into fieldable systems.

Robots are used extensively in manufacturing operations that are repetitive and require precision motions. The automobile industry has been the leader in the application of such devices. Shipbuilding, however, presents a very different set of manufacturing problems. Most ships are fairly dissimilar and built in very small quantities. Many of the structures within a ship hull are modified and changed in the field.

Sensor-based robot systems are therefore indispensable for automating the shipbuilding processes. Several advanced technology programs, such as DARPA's Intelligent Design program and Simulation-Based Acquisition program, have focused on automating the design of ships. The next step in the evolution of this process is to use automated design information to control robots and other devices to perform the manufacturing tasks. Such manufacturing technology has been demonstrated in the past in specialized areas like robotic arc welding. Similar systems for grinding and deburring are being actively researched in the manufacturing community.

Another application of intelligent, sensor-based robots in shipbuilding is in creating as-built drawings of ships by actively surveying ship interiors. This brings together the problems of mobility in restricted spaces, sensing, and interpretation of sensed data into a single application.

One critical and dangerous military task is the handling of munitions, especially on the flight deck of aircraft carriers. Heavy munitions have to be moved from storage to aircraft through the clutter of pipes, hoses, and other obstacles typically found on an active flight deck. These movements have to be coordinated with the other time-critical activity that surrounds the landing, refueling, preparation, and launching of aircraft. Errors in this process can be fatal.

Combining dexterous manipulation and sensor-based navigation and locomotion, munitions-handling systems capable of operating in the flight deck environment (wheeled or tracked vehicles will not be suitable) will provide significant benefits. Further, such a system would be coupled with intelligent processing of logistics and operational-planning information to achieve end-to-end control over the matching of mission-specific munitions to aircraft. The core technologies required to build such a system are already available or are maturing at a rapid rate (see Chapter 10, "Technologies for Enterprise Processes").

The Naval Explosive Ordnance Disposal Technology Division is tasked with the mission of disposal of unexploded ordnance as well as defusing and disposal of improvised explosive devices. This is a fundamentally risky task because it involves putting humans into close proximity to unknown explosive devices. Robotics can play a key part in these tasks.

Robotics technology can be applied to the explosives disposal problem in many ways. The simplest is the use of low-cost, expendable, teleoperated devices to approach, handle, and move explosives to a safe location. This keeps the operator at a safe distance from the explosive device. Teleoperated devices could also be used to neutralize the explosives by using a variety of conventional manual disrupters.

More challenging is the use of robot devices that can make intelligent decisions about the nature of the environment and can position sensors in close proximity to, or even inside, the shell of explosive devices. Such robots require significant dexterity to manipulate the explosives, penetrate the outer shells, and position sensors inside the shell with high precision. Snake-like manipulators being developed today hold the promise of providing such capabilities. In addition, telepresence that conveys high-resolution visual, tactile, auditory, and haptic sense to the operator is crucial because the smallest miscalculation or uncontrolled movement can lead to a catastrophic outcome. Telepresense systems being developed for remote surgery demonstrate many of the key characteristics necessary for explosives neutralization.

The use of unmanned ROVs and their autonomous equivalents (UUVs) provides the naval forces with critical tools for deep-sea search and rescue. High-profile applications such as the search for the TWA Flight 800 debris and photographing of the Titanic have raised the visibility of the technology. This technology is being actively pursued by Department of the Navy laboratories, such as NRaD (e.g., with the Advanced Unmanned Search System capable of searching at depths of 20,000 ft), and by others (MITI in Japan, for example).

In the future, these systems are likely to evolve toward cooperative collections of vehicles. Significant research in core technologies is necessary to achieve this goal. To build an effective collection of collaborative automatons, two things are necessary: each device must be able to (1) model and understand its environment, and (2) communicate this understanding and its intentions to other automatons. Both of these abilities are difficult in an undersea environment because visibility is very limited and communication bandwidth between untethered vehicles is very low, acoustics being the primary medium of effective communications. Both of these limitations place significant stress on the ability of the undersea vehicles to interpret the condition of their environment. Collaboration

between UAVs presents far fewer problems and will probably occur earlier.

The Marine Corps is in the process of evaluating a fundamental change in its operational doctrine. The Sea Dragon doctrine dictates a large collection of small, dismounted units ranging across the target area, controlled from protected cells on offshore ships. Unlike conventional warfare where units are in close proximity to each other and can provide mutual support, the Sea Dragon concept places a significant stress on the small units. Robot assistants can play a significant role in extending the surveillance and offensive capabilities of these small units.

One approach is to develop small, portable, and rapidly deployable robots with a collection of tactically relevant sensors, such as video cameras, audio sensors, and chemical and biological sensors, that can be carried into the battle space by the expeditionary forces. When required, these devices would be deployed to autonomously perform a range of tasks such as watching the perimeter, protecting the rear, and looking around corners or up stairwells in an urban environment. The devices must be autonomous because the marines cannot take on the additional task of teleoperating the devices for any significant periods of time. These challenges are exacerbated by the poor RF communications and difficulties common in urban environments where the Marine Corps will likely operate.

Mines that are originally placed during various internal and border conflicts can remain in place for a number of years. Mines continue to be a very cost-effective defensive weapon and hence are deployed by many adversaries. Of particular interest to the Navy and Marine Corps, therefore, are the detection and clearance of various mines in surf zones, in shallow waters, and in deep waters. Most littoral mine clearance strategies involve the dispatch of Navy Seal units. The consensus emerging in the robotics community is that taskable, expendable machines could significantly reduce the risks involved in such operations for two reasons. First, in view of the inherent wear and tear on the trigger mechanisms of older mines, as well as the movement of these mines resulting from water currents, waves, and other natural phenomena, they could potentially exhibit unreliable behavior; and second, there are areas in the world where sophisticated smart mines have been deployed. Such mines could detect the approach of Navy Seals, explode, and cause lethal damage.

In the past two decades, various programs have been directed at building small, lightweight, low-cost UUVs that can be easily deployed and tasked to deal with various littoral scenarios in deep waters, shallow waters, and surf zones.

The technologies that must be developed in support of these missions are described above.

Automatic target recognition is the use of computers to process outputs of one or more sensors to locate or identify specific objects. The problem is made difficult by the presence of sensor noise, degradation of sensor responses resulting from adverse sensing conditions or countermeasures, and scene clutter. Typical sensors used in ATR systems include forward-looking infrared radars (FLIRs), millimeter-wave imaging sensors, synthetic aperture radars (SARs), and laser radars (LIDARs)

ATR methods themselves fall very roughly into three, heavily overlapped categories as follows:

• Statistical pattern recognition. Statistics regarding some aspect of an object's appearance are used to discriminate targets from nontargets. Many approaches, varying in mathematical rigor and the manner in which features are chosen, fall into the pattern-recognition category. Familiar techniques include Bayesian probability calculus as well as clustering methods.
• Neural networks. Although neural networks are also statistical classifiers, they often differ from typical pattern-recognition approaches in the style of computation and in the means by which features are chosen. Both model-based and statistical approaches require the system designer to choose discriminating features, whereas many types of neural networks can learn features on their own. Some researchers believe that this advantage indicates that the neural networks are the future of ATR and sensor fusion.
• Model-based recognition. Thought to have originated when artificial intelligence techniques were applied to computer vision, the model-based approach relies on a hypothesis-and-test paradigm that attempts to reconcile predicted features based on a target model with extracted features from real sensor data. The extent to which predicted features match sensed data is evidence that a target is present.

When considering factors affecting ATR, one cannot ignore the impact of the following supporting technologies:

• Sensors. Better sensors are generally expected to afford better ATR performance.
• Signal processors. Faster processors mean that more targets, models, and

• situations can be processed. Fuller search and fewer shortcuts will yield better performance.
• Intelligence gathering. The better the a priori knowledge of the target, the better one can design recognition algorithms.
• Target modeling and simulation. Simulation with high-fidelity phenomenology will allow discriminating features to be developed and thoroughly tested.

Naval forces need to hit more targets with higher accuracy and from greater distances using fewer humans. The Department of the Navy is faced with an optimization problem, since it is trying to make the best use of increasingly limited resources. Automatic target recognition plays a key role in this problem, as it essentially replaces human pilots during some or all phases of tracking and homing on both fixed and mobile targets. Improvement in recognition algorithms means that human operators are required to do less and less, freeing them to attend to other problems. Hence, ATR gives the Navy and Marine Corps leverage in solving their increasingly complex resource allocation problems.

Because ATR encompasses a large number of applications and because of the general lack of standardized testing procedures, the current state of the art in ATR is not easily quantifiable. However, some general observations can be made. For instance, there is a widespread reliance on correlation techniques. In addition, variants of correlation, such as the Hausdorf distance—a measure of shape similarity—are also used. An increasingly popular method of integrating fragments recognized by correlation or other means is the elastic template. Regardless of the particular processing steps employed, ATR techniques generally encounter difficulty in highly cluttered scenes. Hence, there is an increasing interest in multisensor fusion technologies to more easily distinguish targets from clutter.

To fully understand future directions in ATR, one must consider the larger context in which these algorithms function. In particular, the success or failure of ATR is dependent on the sensors employed and the a priori modeling used to process the sensor information.

Listed below are trends in sensors:

• Increasing sensitivity, higher resolution, lower noise;
• Large increases in data rates and processing capability;

• Increased use of multiple distributive sensor systems to combat weather and countermeasures and to provide multiple viewing perspectives;
• Increased use of multiple wavelength and multimode sensing to see through foliage and camouflage and to detect low-observable targets; and
• Increased employment of LIDAR and imaging millimeter waves to provide two- and three-dimensional images of targets.

Trends in ATR algorithms are as follows:

• Use of adaptive systems that are insensitive to component failure and changes in target appearance;
• Use of uncertainty reasoning in the representation of ambiguous sensor reports and in the handling of modeling limitations; and
• Increased use of context in processing of models and sensor data.

The increasing sophistication of targets and countermeasures cannot be overlooked. Failure to deal with new countermeasures and stealth techniques may nullify gains resulting from these trends.

The following vision illustrates how ATR could work in the future if the supporting technologies are developed:

• Surveillance missions to supplement data found to be incomplete are automatically performed.
• Target-and-scene models, replete with phenomenology and weather reports, are automatically extracted from existing and newly gathered intelligence databases.
• Relevant target models are communicated to smart weapons and submunitions.
• Intelligence data archives are automatically polled.
• ATR algorithms for on-board smart weapons and submunitions specialize themselves by uploading models, off-board cues, and context information.
• Once approaching the target, smart weapons automatically adjust for differences between the real and hypothesized scene attributes.
• Intelligence databases and simulations are updated from the information sent back by weapons approaching the target and from platforms assessing damage.
• Updated scene-and-target models are unloaded into subsequent rounds of smart weapons and submunitions.

Although ATR can be used in commercial applications, such as automated assembly, typical industrial problems are vastly simpler than those encountered in battle. Hence, the main drivers for improvement of ATR are likely to come from the military rather than the commercial world.

The following developments in supporting technologies and ATR algorithms are needed to make significant improvements in the future:

• Standardized testing through simulations. The establishment of standardized tests and testbeds would allow different approaches and progress in the field of ATR to be objectively measured. No single set of sensor readings can adequately capture the variation encountered in the real world. Hence, rather than evaluating candidate ATR systems based on fixed sets of sensor outputs, simulations incorporating sensor noise and realistic target and countermeasures phenomenology should be used. In simulation-based testing, candidate ATR systems could be judged based on performance in a number of standardized scenarios.
  
  

  The fragmented nature of existing modeling and simulation packages is an impediment to simulation-based testing. The need is clear for an accessible, standardized set of multimode target models. Once available, these models can be incorporated into simulation-based testing procedures for candidate ATR systems.

• Ability to assess information availability. The need to rapidly furnish smart weapons with accurate target-and-scene models is a requirement of the future naval forces' global surveillance and communications mission. In any area of conflict, the naval forces must know what information exists and what information must be collected to create accurate models. The processes of updating intelligence repositories are recurrent, in that the appearance and position of targets change over the course of a conflict. Intelligence repositories must be transparently connected to battle managers and battle managers to weapons and surveillance systems. In the future, ATR will be part of a feedback loop. Precursor surveillance missions having smart-weapon capabilities will seek targets, given the best information of the targets' appearance. Target statistics will be communicated back to the battle manager, and subsequent smart weapons will capitalize on the information.
• Use of flexible recognition schemes. As the use of multiple on- and off-board sensors becomes more common, processing will have to become more flexible in that some sources of information may not always be available. A related problem is that information sources have differing degrees of reliability, particularly in the presence of countermeasures. Future systems must be flexible enough to recognize when the real world is departing from assumptions made during training. ATR systems must be able to measure the quality of various

• information sources and adapt accordingly. For a more in-depth discussion of recognition schemes, see Chapter 3 in Volume 3: Information in Warfare, another of the reports in this series.
• Phenomenology supplanting ad hoc development. In the past, development of many ATR approaches was ad hoc, relying on the experience and intuition of designers for the selection and implementation of discriminators. As countermeasures become more sophisticated and low observables more common, the success of such systems will decline. Instead, a phenomenology-driven approach is needed in which features that can discriminate targets in the presence of countermeasures are automatically or semiautomatically determined. More research in feature discovery and in methods of navigating the mazes of target and sensor physics is needed to accomplish this goal.

Standardized, simulation-based testing could be carried out in parallel within the next 1 to 2 years. Creation of realistic scenarios might require an additional 3 to 4 years.

The information availability assessment task is quite comprehensive in scope. Efforts should build on work done in projects such as DARPA's Battle Assessment and Data Dissemination program, which attempts to provide warfighters with needed information in a timely fashion. After a language is defined to describe what is needed by way of target-and-scene information, an assessment system must be devised to check which pieces of information are already available. Finally, a planning system must be developed to suggest and schedule intelligence-gathering missions. Early versions could simply point out what information is needed and where it can be found. The initial definition phase might require 2 years; a system that relies on heterogeneous databases and inference engines to detect missing and outdated information might require an additional 2 to 3 years. An advanced planning and scheduling system to identify missing information would take far longer—perhaps 5 to 7 years.

Maturation of flexible recognition systems is expected to be a continuing process, taking advantage of more capable sensors and fusion calculi as they become available. Funding should be increased in this area to stimulate continued development and insertion over the next 20 years. Periodic reassessment may be warranted in order to identify and encourage particularly promising areas. Insertion would be a continuing process over the next 2 to 20 years.

Increased use of phenomenology-driven feature selection techniques requires development of tools that simplify the visualization and manipulation of target and countermeasure phenomenology. Funding for such tools and feature-discovery

techniques should be increased and development combined with flexible recognition schemes encouraged. Insertion would continue periodically over the next 5 to 20 years as deep research insights mature.

Automation increases manpower effectiveness and warfighting capability by performing routine functions, conducting superhuman and hazardous operations, and minimizing casualties. The Department of the Navy should field a vigorous program in the technologies for ship automation that will realize these benefits. Unmanned aerial vehicles and unmanned underwater vehicles will play a major role in future naval warfare as surveillance, communications, targeting, and weapon-guidance platforms. The Department of the Navy should support technology developments to increase mission duration and operational capability, enhance sensor payloads, and increase survivability.