IN MAY, the space shuttle Endeavour sent a small cylinder tumbling into orbit
some 280 kilometres above the Earth. It had no onboard computer, no gyroscopes
or other moving parts, not even a set of thrusters. To begin with, the cylinder
spun wildly out of control. But, gradually, the spinning slowed, and within
three days it had stopped.
Today, this brass and aluminium satellite, which is only half a metre long,
steadily orbits Earth. 鈥淚t鈥檚 just sitting there,鈥� says Linda Pacini, the
engineer at NASA鈥檚 Goddard Space Flight Center in Maryland who led the
mission.
Despite this lack of activity, Pacini鈥檚 mission is hugely significant. Most
satellites measure their position and attitude using gyroscopes and
accelerometers, and make corrections with the help of onboard computers and
thrusters. But this satellite is nothing more than a carefully weighted metal
shell carrying magnets that interact with the Earth鈥檚 magnetic field. Called
PAMS (passive aerodynamically stabilised magnetically damped satellite), it is
the first to rely on passive components鈥攖he carefully designed weights and
magnets鈥攖o manage the dynamic process of attitude control.
Advertisement
PAMS is a milestone in the emerging science of passive dynamics: the study of
movement and the way it can be controlled by passive components rather than
active systems hooked up to computers. Passive components are simpler, cheaper
and more rugged than electronic systems. And they are beginning to prove more
versatile at producing some types of movement鈥攚alking, for
example鈥攖hat are too complex for today鈥檚 computers to handle.
Walking robots traditionally rely on active control systems. But robots with
joints and actuators coordinated by computer can rarely negotiate even the
simplest terrain and use up huge amounts of energy. By using legs that rely on
passive components such as springs, weights and pendulums, engineers say they
can dispense with complex electronic control systems and design robots that are
better at walking. The first passive dynamic robots are even now hopping and
waddling around laboratories in the US.
Passive dynamics is hardly a new idea鈥攖he wings on a glider or even the
flights on an arrow are good examples of passive components that control
movement. These were developed in an age when passive components were the only
ones available. In the past twenty years, however, computers have revolutionised
control systems. Modern fighter aircraft, for example, are so aerodynamically
unstable that only a computer can react quickly enough to keep them up. Active
suspensions are gradually replacing old-fashioned springs in cars. And
electronic systems in modern cars routinely take over during heavy braking on
our motorways.
But in the rush to rely on the number-crunching power of computers, simpler
solutions to the problems of controlling movement have been ignored. 鈥淭here has
always been the tendency to choose high-tech solutions because they are sexier,鈥�
says Dave Skillman, the NASA engineer at Goddard who dreamt up PAMS. He prefers
the simpler approach.
The goal of the PAMS mission was to steady the satellite so that it pointed
in its direction of motion. Mirrors fitted to its ends were to be used as
targets for an experimental laser rangefinding system, which would monitor
changes in the cylinder鈥檚 motion. Like any satellite, PAMS is subject to several
small forces, such as the pull from the Earth鈥檚 magnetic field, gravity
gradients and drag from the residual atmosphere. 鈥淭hese forces have
traditionally been a problem with almost every satellite and a lot of work is
put into subduing them,鈥� says Skillman. The trick is to use these forces to your
advantage, he explains.
Turning somersaults
For example, Skillman and another NASA engineer, Tom Flatley, used a
magnetically 鈥渟oft鈥� alloy, called AEM 4750, to harness the Earth鈥檚 magnetic
field and stop the satellite tumbling after its ejection into space. When placed
in a magnetic field, each atom or group of atoms in the iron-nickel alloy acts
like a small magnet and aligns itself with the external field. If the external
field is reversed, the field inside the material changes too. But during the
time the internal magnetic field takes to become realigned, any residual field
in the original direction opposes the changing field.
PAMS carries three wires of this alloy, each 30 centimetres long and aligned
with each of the cylinder鈥檚 axes. Because the satellite is somersaulting through
space, the direction of the Earth鈥檚 magnetic field relative to the wires
constantly changes. And as the magnets struggle to realign their internal fields
to match, they create a turning force, or torque, that tends to oppose the
tumbling.
The torque created is roughly one micronewton-metre鈥攁bout a hundred
billionth of the torque needed to tighten a bolt. But it is enough to convert
uncontrolled somersaults to an acceptable 7 degrees swing in about sixty orbits,
or three and a half days. 鈥淭he magnets are never really happy, so the satellite
will never be completely stable,鈥� says Skillman. 鈥淭hat鈥檚 one of the
迟谤补诲别-辞蹿蹿蝉.鈥�
But with the tumbling under control, another effect becomes significant.
Satellites are vulnerable to an odd phenomenon called the gravity gradient
torque, which is caused by the fact that the force of gravity falls with
increasing distance from the Earth. Because of this torque, satellites tend to
become aligned with the gradient. 鈥淓ven a pencil orbiting the Earth would tend
to become vertically aligned,鈥� says Skillman.
Deceiving gravity
Of course, an orbiting sphere would not experience this torque because its
mass is distributed evenly around its centre of gravity. The crucial property of
a sphere that gives it immunity to the gravity gradient torque is its moment of
inertia (its resistance to rotation). For a sphere, this is the same regardless
of the way in which it is spinning, but most other objects have moments of
inertia that depend on their axes of spin. For instance, a dumb-bell is hard to
twirl like a baton but easy to roll along the floor. The trick that Skillman
employed was to distribute the satellite鈥檚 mass so that its moment of inertia is
the same in all three axes. 鈥淲e turned it into the mass equivalent of a sphere,鈥�
he says.
To achieve his final goal鈥攖o point the satellite along its direction of
travel鈥攈e turned to conventional aerodynamics. At an altitude of 280
kilometres, the upper reaches of the atmosphere produce a small but significant
amount of drag. The force this creates is roughly 100 micronewtons鈥攁bout
the weight of a staple, or forty times the force that sunlight exerts on a
satellite. Skillman used this drag to control the attitude of the vehicle.
The same principle ensures that a dart always points in the direction of
motion. This is partly because of its shape, but also because most of its mass
is concentrated in the nose. 鈥淚f you put a dart into a cardboard toilet-paper
tube and sealed off both ends, it would still fly nose-first,鈥� says Skillman. In
this case, both ends of the tube feel the same drag, but this force has less
effect on the more massive end.
PAMS is similar鈥攁 cylinder with most of its mass concentrated at one
end so that its centre of mass is only 8 centimetres from this end. If the
satellite were the size of a beer can, says Skillman, it would balance 鈥渉alf an
inch鈥� from one end. Like a dart, the massive end of the satellite points into
the atmospheric wind. 鈥淚t鈥檚 a very elegant device, being as passive as it is,鈥�
he adds.
During the mission in May, the satellite was stable and pointing into its
direction of motion within three days of its release. In future, Skillman
believes that passive dynamic satellites will be used as targets for laser
experiments in space. These experiments could lead to a new generation of
communications satellites that rely on lasers rather than radio links.
The satellite鈥檚 biggest limitation, however, may be NASA itself. 鈥淧eople in a
high-tech culture expect high-tech solutions and what we did with the satellite
was decidedly low-tech,鈥� says Skillman. 鈥淧eople kind of look down on it.鈥� They
can鈥檛 argue with the price鈥攖he total system cost $600 000. 鈥淚n
shuttle terms, extremely cheap,鈥� he says.
Robotics is another high-tech field that has been slow to embrace passive
dynamics. Science fiction, for better or worse, has spawned visions of robots
folding laundry and bringing the afternoon tea. Engineers, however, would be
content with robots capable of walking across rocky ground.
Efforts to date have been unsuccessful. Traditionally, engineers have tackled
the problem using robots that sense and control the position of every joint in
their bodies and legs. Such an approach requires powerful computers to process
the constant stream of data from the sensors and calculate how to operate each
joint in a way that moves the robot forward. A typical system might make 100
measurements and calculations per second. But even robots using the best
algorithms have trouble negotiating anything other than smooth terrain.
The human body, however, makes do with fewer signals. 鈥淭he fact that the
human body does it better is a clue that we must be going wrong somewhere,鈥� says
Jerry Pratt, an engineer in the Leg Laboratory at the Massachusetts Institute of
Technology, in Massachusetts, where researchers are working on a different
approach.
Their ideas date back to 1989, when an engineer named Tad McGeer built a
remarkable pair of metal legs that relied on passive dynamics rather than
complex computer control. McGeer鈥檚 legs are 90 centimetres tall, bent at the
knee and fitted with paddle-like feet. They have the same proportions as human
legs and freely rotating joints. Unlike other walking robots, they have no
motors or sensors and no onboard computer. But when placed at the top of a ramp,
they walk downhill with a rolling gait that is strikingly human, like a cowboy
sauntering towards a high-noon showdown.
Once on the ramp, the legs鈥� centre of gravity tends to make them topple
forwards. This shifts the weight onto the first or forward leg, which acts like
an inverted pendulum swinging the hips forward. During this motion, the hips
rise, lifting the trailing leg, which swings forward. And as the hips descend,
the foot of this leg lands on the ramp in front of the first. This then becomes
the pivot of an inverted pendulum. As the robot topples even further, the hips
rise, lifting the trailing leg and so on. The steps continue until the robot
reaches the bottom of the ramp.
Although McGeer鈥檚 legs do nothing more than walk downhill in a straight line,
they prove that this ability is an inherent property of certain physical shapes
and structures. Walking is a dynamically passive activity.
For all their elegance, McGeer鈥檚 legs can only walk downhill. So the
challenge in 1989 was to find a way to apply passive dynamics to more useful
robots. At about the same time, a robotics engineer called Marc Raibert, who
founded the Leg Laboratory and brought it to MIT, was working on ways to build
simple, cheap robots that could negotiate more complex terrain. He immediately
spotted the potential of passive dynamics.
His approach was simple yet effective: he added springs to his robots鈥� legs.
鈥淚t was clear that animals used springs,鈥� he says. 鈥淚t seemed conceivable that
robots could make use of the same mechanism.鈥� The muscles and tendons in the
legs of kangaroos and rabbits act as springs that stretch and store energy.
Raibert found that springy, pogo-stick legs allowed his robots to bounce and
balance at the same time, with only a modicum of help from a computer. With just
a little more help they could negotiate tricky obstacles such as stairs and
puddles.
Wandering robots
Raibert combined passive dynamics with simple computer programs to build one,
two and four-legged robots that could hop, run, climb stairs and jump through
flaming hoops. His results were groundbreaking. 鈥淢arc鈥檚 work ultimately changed
the way people think about robotics,鈥� says Matt Mason, a computer scientist at
Carnegie Mellon University in Pennsylvania. 鈥淧eople are much more broad-minded
now about robotics and what are valid approaches.鈥�
But there has always been one major problem. Raibert鈥檚 machines used vast
amounts of energy. His robots had to be tethered to powerful hydraulic pumps, so
they could never wander far from home. Autonomous robots must use energy
efficiently enough to survive with only small electric motors and batteries on
board.
While energy-efficiency was not Raibert鈥檚 goal, it may be his legacy. Martin
Buehler, an assistant professor at McGill University in Montreal who studied
under Raibert, and his student Mojtaba Ahmadi, have built one of the most
energy-efficient legged robots to date, a one-legged hopper that moves at up to
1.2 metres per second and uses about 125 watts. Buehler hopes to cut energy
consumption further by using a passive dynamics technique first proposed by
Raibert in 1989. It involves a single spring
Buehler鈥檚 one-legged hopper resembles a shoe box on a pogo stick, with a
single moving joint which swivels at the hip. It has two motors: one in the leg
to make it hop and another in the hip which swings the leg forward during each
hop. Each motor provides about half the robot鈥檚 total energy requirements.
The trick that Raibert suggested but never tried was to add a spring to the
hip joint that would help the leg return to the starting position after each
hop. Buehler is now working on this refinement. The strength of the spring and
the mass of the leg govern the leg鈥檚 natural swinging frequency, which is
similar to the natural frequency of a pendulum. When the swinging rate is
rhythmic, the robot remains stable and the spring, rather than the hip motor,
does most of the work.
But keeping a freely swinging leg under control is difficult. So the robot
has a sensor that monitors the hopping rate and a computer that analyses the
data. Whenever the leg deviates from this rate, the hip motor kicks it back into
its natural rhythm, like nudging a pendulum to keep it going. The algorithm
needed to do this is remarkably simple. In computer simulations, the hip motor
uses far less power, cutting the total energy consumption of the robot by a
quarter. 鈥淪aving energy in the hip swing will go a long way toward making legged
robots autonomous and practical,鈥� says Buehler. If all goes to plan, he should
have the improved robot running by the end of this month.
Passive dynamics is set to revolutionise the design of other robots too. At
IS Robotics in Somerville, Massachusetts, engineers are building some of the
most complicated robots in the world. For now, the company鈥檚 robots use no
passive dynamics. 鈥淓nergy consumption is always a concern,鈥� says Helen Greiner,
vice-president of engineering at the company. Buehler believes that passive
dynamical systems can save substantial amounts of energy. Greiner agrees: 鈥淚t鈥檚
certainly something we would have an interest in.鈥�
Today, the greatest challenge, whether in space or on the ground, is
convincing scientists that passive dynamics can lead to simpler, more efficient
and reliable machines. Despite the optimism of researchers like Skillman and
Buehler, they are in a minority. Whether other scientists can be converted to
the cause remains to be seen.