LEVEL 1 - 2 OF 115 STORIES
Copyright 1994 Jane's Information Group Limited,
All Rights Reserved
International Defense Review

November 1, 1994

SECTION: Vol. 27; No. 11; Pg. 55

LENGTH: 3294 words

HEADLINE: MARCH OF THE INSECTOIDS

BYLINE: Artur Knoth

 BODY:
    The new generation of small autonomous land-based robots has a wide
application range from reconnaissance to weaponry. Artur Knoth reports.

    As IDR has noted before, the voting public in many Western nations, shocked 
by graphic television coverage of recent wars, is opposed to unnecessary
collateral damage among civilians and to high casualty figures among its own
military personnel. This is encouraging the development of both nonlethal
weapons (NLW; see IDR 7/1994, pp.28-39) and of unmanned, remotely controlled
vehicles.

    A new concept for landbased military operations using small autonomous
vehicles or "insectoids" has weapons applications which cover both the lethal
and NLW sectors, but also reflects trends in civilian technology and parallel
developments already under way for air and spaceborne systems.

    Such robot systems are smaller, less complex and cheaper than their manned
equivalents. The much smaller signature reduces vulnerability, increases
survivability and allows the robot to get closer to the target, resulting in
higher accuracy and precision. While extremely important for penetration behind 
enemy lines in land battles, this feature also has a political value: objections to
 high collateral damage have raised expectations as to the possibility of
"surgical" target attack. This, however, will require better intelligence
capabilities than are presently available: reconnaisance needs to be continous, 
comprehensive and on or near site for true surgical strikes. Snapshots via
satellite or drones are not timely enough.

    Insectoid developments parallel trends in other technology sectors: for
example, various space programs have proved that systems can become too large
and complex and, if one part fails, the whole expensive system may be rendered
useless; where the possibility exists, therefore, such systems are now broken
down into smaller, cheaper modules or independent overlapping units.  Hence, if one
satellite fails, the rest of the constellation can continue to function.  Meanwhile,
in computer technology, each system tends to be "only as smart as really necessary,"
in order not to hold up processing because of overly complex controller architecture.
This principle is now also applied to AI (artifical intelligence) used in robotics.

    New approach

    The ALV (Autonomous Land Vehicle), part of ARPA's (the US Advanced
Research Project Agency) strategic computing program, is typical of much of
the activity in military robotics hitherto. The ALV was a small truck or van-sized
vehicle that could navigate a well-marked road at a few miles per hour. An
optical system, together with the required computing facility, performed the
classic, step-by-step approach to its target - detection, determination of
orientation, recognition and identification. A backup digitized map was stored
in the computer, in case the primary system failed, and a sequential process was
used to determine the commands needed to keep the vehicle on the road. Sensor
input was analyzed and compared, and the course of action calculated with the
architecture demonstrated on the left hand side of the figure on p.56. Other
versions, such as the UGVCT (unmanned ground vehicle control testbed) are
really telerobots, steered by a controller in another vehicle and thus not
strictly speaking autonomous.

    A better approach is shown on the right hand side of the figure on p.56. In
subsumptive architecture, as it is called, very simple commands form the
elements, such as:

    - Climb over obstacles;

    - Look for shadows;

    - Seek sources of noise;

    - Avoid touching a wall, etc.

    Each command uses the sensor data and the ultimate course of action depends
upon the weighting (or rank, priority) given each individual instruction.
Initial versions of small robots based on this minimalistic approach were able
to "teach" themselves to "walk," even though all six insectlike legs were
independent of each other. The subsumptive approach carries within it the basis
of a neural network.

  Examples of the prototypes already afoot in the laboratories include:

    - Allen (created by a team at the Massachusetts Institute of Technology).
Basically a footstool on wheels, this demonstrated how the use of sonar readings
for the command layers allowed the robot to cruise around the room avoiding
obstacles, following the walls and finding doors and to exit via the centre of
the nearest doorway.

    - Genghis (also MIT), a 35cm-long robot with six simple legs that can track
people with its infrared "eyes" and has 12 behaviour layers of instructions. An
improved version, Genghis II, costs only some US$6300 to build.

    - Squirt (MIT), a 3.3cm "gnat" robot that seeks out the dark and hides,
venturing out only to investigate loud noises. This is achieved with a mere 1300
bytes of computer code for the control system.

    - Attila I (MIT) weighs in at 1.6kg, has 23 motors, 10 microprocessors and
150 sensors, and also carries a miniature video camera. Attila is able to climb
over obstacles, as the six legs are in sockets that allow more freedom of
movement than for Genghis.

  - Herbert (MIT), with a mechanical gripper arm and laser vision, roams around the
laboratory and is said to steal empty soda cans.

    - Robotic Solar Mower (Outdoor Productions, Shreveport, Louisiana, USA) is a
solar-powered lawnmower that, once the wires that delineate the boundaries of
the area to be mowed are buried, mows the lawn completely autonomously, even
avoiding collisions with pets and other obstacles. The unit sells for US$2000.

    - Khepera (Lausanne university) weighs only 70g. It has a gripping arm, two motors,
two CCD cameras and an infrared distance-measuring device. In this case, "open"
programming with a neural network was used.

    For any insectoids to be able to fulfil militarily useful tasks, they need a broad
spectrum of technologies with which they can be equipped. Some enabling
technologies now available, and in part due to continuing miniaturization, are:

    Multiple chip modules (MCM): joining several chips together into a module
opens up new possibilities of performing a wider range of tasks, quite apart
from performance increases in the indvidual chips themselves.

    Sensors: for imaging (visible and IR); fibre-optics, photonics and
electro-optics are creating high-performance sensors and cameras that are so
small and lightweight that, as indicated, insectoids can already carry a
versatile sensor package. For example, the current new generation of diode
lasers on a chip could enable insectoids to perform many tasks, even though they are
extremely small. Applications could range from acting as a mobile laser
radar to detecting chemicals in the air. Besides imaging sensors, others are
available. The spectrum ranges from accelerometers, magnetic and/or electrical
field gauges to acoustic sensors or silicon "noses" - sniffers to detect gases
and other chemical agents. Due to the progress in micro and nanomechanics, many
of these can be integrated on to an MCM. Several different types can be combined
into a single MCM package.

    Communication: considering the small size of the newest mobile telephones
and the availability, in the future, of large satellite communication systems,
insectoids could even have limited communication facilities on board. The link
would make use of the latest encryption and message-compression (burst)
techniques.

    Further applicable technologies are in the navigation, energy and data-processing
areas:

    GPS: with GPS chips becoming ever smaller, an insectoid with a GPS receiver on board
would be able to "know" where it is at all times. Insectoids equipped with GPS could be
programmed to remain inside clearly defined geographic limits.

    Solar cells: work on thin-film amorphous silicon photovoltaic solar cells
has succeeded in creating a version that has an efficiency of over 10% and is
deposited on flexible stainless steel sheets. This allows the cells to be placed
conformally on objects, including curves and bends. In this way the whole
structure of the insectoid is available to gather energy for the batteries.

    Batteries: some of the advanced lithium batteries consist of extremely thin layers.
Their high energy density coupled with the fact that these sheets too can be shaped to
conform to the body of the platform, allow lightweight systems where the batteries
and solar cells actually become a part of the structure.

    Data storage: new photorefractive materials (also known as spatial light
modulators, or SLMs), are ideal data-storage devices. SLMs could enhance the
intelligence of the insectoids (especially with regard to the holograms needed
for advanced pattern-recognition), without exacting a too-high weight and/or
volume penalty. Even presently available systems are adequate to meet current
needs.

    Data processing: Neural networks have already been introduced and should be more
widely used. Furthermore, more effort should be invested in the correlation (that is,
fusion) of the output of the different sensors. Such correlations permit identification of
objects, even when all the sensors are not high resolution. Studies should determine what
values of low-resolution magnetic, acoustic and optical sensors taken together can
identify a tank as opposed to, say, a bus.

    Insectoid systems with military applications can be grouped into four major categories:

    - Site security;

    - Intelligence (including covert);

    - NLWs;

    - Lethal weapons.

    Equipped with passive (such as optical and acoustic) and/or active (sonar,
etc.) sensors, insectoids could be used instead of sentries or guard dogs. For
fixed sites such as bases or research facilities which already have one or more fences,
insectoids could be programmed to follow the fence and never stray
further than a fixed distance from it. In the case of two fences, the
insectoid guard would move between them. The robot lawnmower cited above would
provide a model for such a system.

    In the field, for site security of a temporary nature, insectoids could
reduce or eliminate the guard duty undertaken by humans, especially when a unit has
several sites over a large area. One way to set up such a system would be
that demonstrated in the figure above right, using a beacon system.

    Security for sites scattered over a large area, such as missile batteries
with a central control, could be undertaken by insectoids programmed, via an
intensity-gradient sensor (radar, sonar etc.), to remain between two values of
the emitter field. When the field intensity reaches either the lowest or the
highest programmed value, the insectoid turns around. The same can be done with
wires or portable fences and the architecture need not be closed, as the figure on
 p.56 shows, but can accommodate partial screening. A neural network, with the
ability to learn and which could make use of threshold values derived from
experience, could help to keep the false-alarm rate very small.

    Self-defense armament would be unnecessary in the case of insectoids, but if
the intruder is to be detained, then arming the insectoids with nonlethal
capabilities (NLW discussed below) must be considered. Insectoid sentries could a
lso be armed with flares, percussion caps and other devices which would deter
intruders and at the same time signal their presence, robbing them of the
critical element of surprise. Further roles for insectoids as roving on-site
"watchdogs" can be envisaged with regard to verification of arms-control
agreements and for "policing" facilities where proliferation activities are
suspected.

    Gathering intelligence on the ground behind enemy lines remains a problem:
systems like AWACS and JSTARS provide "standoff" intelligence but have inherent
problems of resolution and discrimination; intelligence drones, unmanned aerial vehicles
and airborne reconnaissance cannot deliver a continous picture of events in any given area:
the information obtained is always just a snapshot view. Furthermore, such
intelligence-gathering systems are prone to betray their position as much as any manned
foray, yet the the best intelligence is that obtained when the enemy is unaware that he is
being observed.

    Here, insectoids could indeed fill a gap. Equipped with the necessary sensors
uite (which can differ from case to case), GPS and mobile communication, an
insectoid could be instructed to patrol a certain area (denoted by the stored
GPS co-ordinates) and report whenever something happens within that area that
fulfils the criteria loaded into its memory. Being small and very flexible,
insectoids may not attract enemy attention; even if one or more are discovered, the
enemy cannot be sure he has found them all.

  In theory, such insectoids can cover quite a wide range: for intelligence
required from close to the front, they could just "walk" in. For intelligence
from deep inside enemy territory, they could be airdropped. As they are very
small and lightweight, they could be released from a cruise missile, drone or
manned aircaft during flight. If placed in spherical shells equipped with small parachutes,
they would either not be noticed by enemy warning systems or would
be interpreted as decoy or dud rounds. By the time their true nature has been
detected, the insectoids should already have scampered away to gather
intelligence. In the 1990-1991 Gulf War, for example, such insectoids could have hidden in
the sand dunes or among the rocks, continously monitoring activity at areas where mobile
Scud-missile units were suspected.

    In peacetime, insectoids could help prevent adventurism such as Iraq's
invasion of Kuwait, and could play a major intelligence role if employed against perceived
threat countries like North Korea or Libya. The danger of an armed
conflict breaking out over a few dozen insectoids, if discovered, would not seem very
great. The value of insectoids in counter-proliferation activities would be
high, as their sensor suites could be selected to cover NBC weapons, and all
kinds of missiles of concern, as well as conventional arms. The placing of such systems
should not be beyond those who travel regularly and legitimately between countries.

  Insectoids equipped with NLWs would be stationary or roving, depending upon the
 situation and task at hand. Used for security or for access denial, rather
like a mine in warfare, they could be employed on peacekeeping or peacemaking
missions. Particularly for this application, a GPS capability to keep the
insectoid from straying too far, and a communications facility, to enable recall of the
insectoids to base, is important. The weapons would also depend on the
mission: for site security, strong light and/or sound effects could suffice, as indicated
above. The other NLW possibilities available are limited, due to the
insectoid's small size; however, canisters of tear gas or other non-harmful
chemical sprays, glues or lubricants could be considered. It should also be
possible to spray a marking dye that allows later identification of the target. Other
"weapons" could be low-power EW (electronic warfare) and ECM (electronic
countermeasure) jammers for specific localities.

    Public distaste over collateral damage has also led to a United Nations
proposal to ban most land mines. Since these are an important weapon in the
military arsenal, a new approach is called for in designing the next generation of area
access-denial mines. The larger lethal payloads that could be considered
for insectoid carriers include a new type of mine which does not endanger the
civilian population to the extent that current designs do. These would be
"intelligent," but in a quite different way from current intelligent mines.
Their very mobility and the fact that they can hide would mean that these
loitering, autonomously terminally guided machines would be unlikely to be
spotted by mine-clearance units.

    The basic intelligent-mine payload of a charge, a fuze, and a sensor system to
determine when, in what direction and against which target it is to detonate will not
change much. However, as the insectoids are mobile and seek out their
target, their munitions load could be lighter than the bulky shaped charges
presently used. This is because, for attack on an armoured column behind the
front, for example, disabling a lead vehicle by blowing off a tread or
destroying a tire could suffice to stall the advance. Repeated attack will
ensure that spares will soon run out, resulting in the halting of the column
until logistics has caught up. Furthermore, a multisensor suite making use of
the correlations and neural networks discussed above will require much less
memory and contribute less of a load.

    The GPS capability of an insectoid mine would enable it to be programmed to monitor a
specific area and/or target (such as crossroads, bridges), but it would not be fixed in
place. An on-board receiver would allow insectoid mines to be repositioned to new GPS
co-ordinates via radio link. This would also allow disarming or even recall of the mine on
cessation of hostilities, thus decreasing the likelihood of injury to the civilian population
after the war.

  An additional safety feature would be a failsafe tag which would allow the
mines to be safely de-activated or at least located if the communication link is
broken. The kind of semi-active device envisaged for arms-control verification
applications (see IDR 5/1991, 411-413) would be suitable, namely one which
responds when challenged in code, using either a battery or the received
challenging signal as the source of energy for the answer. If clever coding is
used, the tag will not give the enemy an opportunity to convert this into an
effective countermeasure.

    These mines could be airdropped near a target, then walk up to it, opening
up the prospect of their use as an anti-infrastructure weapon. Typical targets
would be bridge supports, pipelines, substations, railway tracks and power
lines. In the last case, for example, the charge could consist of the carbon
fibres used against Iraqi power plants in the 1990-1991 Gulf War and delivered
by Tomahawk missile.

    Finally, their use as a counterproliferation weapon can be foreseen (see
IDR10/1994, pp.31-48): where proliferation has already occurred and an embargo
has proved useless, insectoids could be used as covert weapons against suspect
facilities. Armed with conventional explosive, shrapnel rounds or even fuel-air explosives,
the insectoids would attempt to penetrate as far as possible into
the facility before detonating; if discovered and tampered with, they should
also produce the same effect.

    Notes 1. See R.A. Brooks: "New Approaches to Robotics"; Science 253 (5025), 1
227 (13 September 1991).

    References R.A. Brooks (see note 1) T. Beardsley: "Hot Property"; Scientific American 270
(4), 90 (April 1994) D.H. Freedman: "Invasion of the Insect
Robots"; Discover 12 (3), 42 (March 1991) R. Mestel: "Here's Looking at You,
Genghis"; New Scientist 138 (1872), 18 (8 May 1993)


     Insectoid categories and the tasks they could fulfill:




             Security     Intelligence      NLW      Lethal
                                          weapons
Peacekeeping      *        *                 *
Peacemaking       *        *                 *           *
Armed conflict    *        *                 *            *
Quasi conflict     *        *                 *
Proliferation      *        *                 *
Verification       *        *                N/A       N/A



GRAPHIC: Photograph 1, Genghis has six infrared eyes with which it can track,
for example, humans; it has 12 behaviour layers of instructions.  (photo: R.A.
Brooks, MIT); Photograph 2, Attila, a Massachusetts Institute of Technology
creation has six legs socketed so as to allow more freedom of movement than for its
predecessor Genghis and which enable it to climb over obstacles. (photo:
R.A. Brooks, MIT); Diagram 1, A neater approach Only as smart as really
necessary: computing became cumbersome when too-complex architecture began to
hamper processing and as fast calculation became more important. Reducing
decisions to a minimum by means of the Reduced Instruction Set Chip (RISC) has
an analogous trend in robotics: the temptation to make the robot smarter than
necessary for the task is avoided, by means of subsumptive architecture
pioneered by MIT.1 In the figure, traditional, sequential architecture is shown at left and
alternative, subsumptive architecture at right.; Photograph 3,
Squirt hides in dark places and emerges to investigate loud noises. (photo:
Brooks, MIT); Diagram 2, Beacon-directed insectoids establish a secure area in
the field