Paradigm Shifting Without A Clutch
Marriage Peril

by Douglas PageŠ1997

Mark Tilden, the father of BEAM robotics, read his first book and built his first robot before he started kindergarten. Now, because so many elaborate satellites fail, he is designing remarkably simple spacecraft so small they would fit in your hand. "Maybe it's paradigm shifting without a clutch, but at least our devices work," he says. 

Mark Tilden's future was bumped out of its ground state as a toddler. He built his first robot at the age of three. A year later he read his first book, "Way Station", by Clifford D. Simak (Tachyon, San Francisco, 1961).

"Way Station" remains Tilden's favorite book. "It details the wealth of the human spirit emphasized, not absorbed, by technological advances. It's something I trust my machines might someday assist in." Thirty four years later, his interest still in an excited state, he is widely acknowledged as one of the world's foremost experts in robotics.

Like most robot builders, Tilden at first tried building complex machines. He wanted one to vacuum his apartment. But, after loading thousands of lines of code into the robot's memory to anticipate even the smallest change in the machine's environment, Tilden discovered he had succeeded only in making the machine paranoid. The contraption would vacuum an empty room, but once coffee table legs, a few vagrant socks, sneakers and wandering, indolent cats were added it became impossible for the xenophobic robot to do much more than swivel in uncertain circles.

Then, in 1989, Tilden attended a lecture by Massachusetts Institute of Technology professor Rodney Brooks, who described the idea of building robots without processors for brains, instead relying on reactive sensors attached to motors, in effect reducing the complex to its most fundamental constituents. Tilden had his next robot - the size and elegance of a pencil cup - constructed of a couple of transistors and solar cells borrowed from a discarded pocket calculator within two weeks. When he flipped the critter over, to his surprise it started moving toward the light.

Since then the parturient Tilden, whose favorite character from literature is Homer Simpson, has assembled over 300 "biomorph" creatures - walkers, hoppers and crawlers of over 35 "species", with reactive sensors acting as central pattern generators, which have proven reliable and nearly impervious to electrical and mechanical failure. Plus they seem surprisingly capable of collective behavior. They all work on a simple control loop - he calls them Nervous Nets - that sends messages from one leg to the next to keep the creature moving.

Tilden, born in Shroud, England, the gifted son of a housewife and a salesman, then reared and educated in Ontario, Canada, found he could create these insect-like robots based on what he calls Nervous Nets (Nv) and BEAM principles (for Biology, Electronics, Aesthetics and Mechanics) with remarkable survival skill. Tilden's Nervous Nets work like the neurons in animal nervous systems, which put out spiked pulses that hold useful information in the timing between the pulses. In computer terms, he says, the information needed to do work lives in the firing rate, not in a coded voltage level. His creatures can be assembled using as few as two 'neurons'. With a slightly larger array of neurons, Tilden erects sophisticated walkers controlled by central patten generators that are directly coupled to their environment.

Nv technology is based on the notion that a brain is of no use without a spinal cord. Indeed, most life on earth survives just fine without any brain at all. A nervous net, according to Tilden, is a non linear analog control system that solves real-time control problems normally quite difficult to handle with digital (microprocessor) methods. Nervous nets are to neural nets the same as spinal systems are to the brain, he says.

The basic principle of BEAM is to forget the brain and focus on a simple stimulus-response based ability within a machine, says Tilden, who left work as a robotics and computer engineer at his alma mater, the University of Waterloo in Ontario, in 1993 in favor of a position as research scientist in the Physics Division at Los Alamos National Laboratory, New Mexico.

In it's purest iteration, a solar powered BEAM creature uses minimalist electronics from recycled and reused techno-scrap components, such as calculator solar cells, transistors from portable radios, motors from old toys and volt meters.

"The science behind the idea stems from current concepts in artificial intelligence, artificial life, evolutionary biology and genetic algorithms," says Tilden. "Building large complex robots hasn't worked well, so why not try to evolve them from a lesser to a greater ability as mother nature has done with biologics?"

The simplest example of BEAM robotics can be constructed around a two-transistor Nv circuit, called a relaxation oscillator. The results are simple, elegant machines with no processors involved. The most surprising thing about these non-linear crawlers is not that they can out-maneuver other machines, it's their parts count. The most fundamental biomorph creature consists of a relaxation oscillator core consisting of six Nv elements, each equivalent to two transistors. This 12 transistor Nv is sufficient to control and maintain a steady walking pace in many biomorph designs.

"It has been demonstrated that these systems, when fed back into themselves rather that through a computer-based pattern generator, can successfully mimic many of the attributes normally attributed to lower biological organisms," says Tilden. "Using Nv, highly successful legged robot mechanisms have been demonstrated which can negotiate terrains of inordinate difficulty for mechanisms with wheels or tracks."

In the ensuing years, BEAM robotics has become so popular among 'bot builders that its own BEAM Robotics Games, established by Tilden in 1992, now attracts entrants from around the world to Albuquerque each year to compete in 14 events from solar rolling to rope climbing.

Recently, Tilden and LANL colleagues Kurt Moore and Janette Frigo have taken these minimalist BEAM principles to a higher level. They're creating constellations of highly compliant microsatellites, designed to survive in space without struggling to anticipate and overcome the harsh environment there, machines that "negotiate" rather than "bully" their way, that both move and measure using the same actuators.

Trend in Unmanned Spacecraft

Little by little, the trend in unmanned spacecraft design has evolved toward smaller, faster, cheaper. We no longer lift so many high-performance Bentleys and Lamborghinis into orbit. Low earth orbit now contains a number of Ford Escort-class vehicles.

It's not just the stifling budgets. Too many of the costly, complex satellites launched recently have failed. The more complex the designs the more prone they seem to be to failure. Balky antennas refuse to deploy. Flashes of space radiation burn-out delicate microprocessors. Links with earth control stations are lost. This vulnerability has lead some scientists to rethink traditional design philosophy that attempts to predict every anomaly and then pre-program the appropriate circumvention.

"We can't anticipate everything that can go wrong, especially in space and especially under current cost-risk constraints that bind space programs," says Tilden. "We need engineering solutions that respond to unanticipated events in non-catastrophic ways." In other words, everything has to be done differently. Tilden thinks even the Escort-class satellites are too complicated. He'd like to fill parking orbits with Hot Wheels.

What Tilden is suggesting is a paradigm shift away from the highly evolved space systems currently in operation. Rather than design spacecraft systems primarily to perform work and hope that they survive all anticipated problems, Tilden has turned design conception around. "We must instead design systems that automatically attempt to survive all circumstances and then try to extract useful work," he says.

"They represent the logical end of one evolutionary trend in space exploration toward clusters of small, cheap satellites that can achieve results even when some of them fail," says LANL scientist Kurt Moore. "They're so robust they could even be used to investigate the Van Allen radiation belts, which most earth-orbiting satellites avoid."

Instead of designing systems that withstand their environment, the LANL team is engineering systems that rely on their environment. "We're working on satellites that have no microprocessors or fixed algorithmic behaviors. Instead, these satellites are survivors - designed from the bottom up and domesticated by their sensors and control payloads into performing high-reliability tasks," says Moore.

If smaller is better for survival, how small can you make a satellite? At the December, 1997, American Geophysical Union meeting, the LANL team presented an idea for swarms of microsatellites, each weighing no more than half an ounce with control systems based on the simplest Nv "twitches". You could carry these satellites in your pocket and launch them out of the shuttle cargo bay with a slingshot.

Their "satbot", resembling a carefully folded sheet of construction paper, has three solar cells, a long range antenna dish the size of a button, a short range antenna probe no longer than a toothpick, a local sensor array and three alignment coils. It weighs 10 grams and would fit in your hand.

This weight not only reduces cost, but improves accuracy and response of the satellites for ranging and orientation purposes. The basic idea is to launch these satbots by the hundreds, huge stacks of them shearing themselves in orbit as a sweeping array of sensory "pixels", giving a whole new spectrum on the potentials of data measurement.

Short-range, local antennae establish communications between local satbot "cells", while long-range antennae handle distant signals.

The Nv prototype satbot control system uses gradients in the earth's magnetic field as sensed by its magnetic alignment torque coils to stabilize itself and simultaneously uses light gradients from a photosensor pair to orient the spacecraft toward the sun. The advantage is, since Nv controllers are asynchronous, the satellite can match the complexity of the environment at the rate its sensors can perceive. The researchers say this allows the machine to automatically adapt to the stress of its condition, a biological feature that can sometimes produce more constructive behavior than just waiting in orbit like a rock.

The whole microsatellite idea is founded on the fact that space exploration is expensive and risky. However, Tilden says, satellites can be built cheap, rugged, self-programming and "rad-hard". These microsats may be the first step to the potential of a whole range of long term, low cost space missions that would otherwise be prohibitively expensive.

Beyond mapping the Van Allen Belt, other potential applications include measuring the solar wind and more precise earth imaging. That's just the beginning.

Low Earth Orbit Debris Sweepers

"We've done work on everything from lunar dust clearers, lunar miners and sorters, self-organizing heat-resistant tiles, broad-field pixelsats that can image asteroids or vast areas of space to low earth orbit debris removers and single mechanism interstellar explorers," says Tilden. "And there's not a computer in any of them. Most of them aren't even as complex as a transistor radio. Maybe it's a paradigm shifting without a clutch, but at least our devices work."

Tilden's long range goal is to set up a whole organized branch of alternate robotic sciences powerful enough to think about colonization, and not just interstellar exploration. "The benefits will come from cheap access to space for people other than the big boys, and in innovative techniques that don't require NASA dependence. Find a way to sell a Sharper Image satellite for a thousand dollars with your choice of small payload, accessible and/or controllable through the Internet and you'll see a revolution in amateur astronomy equal to the first ground glass lens."

The Los Alamos prototype is a microsatellite whose primary mission is to orient itself in earth's magnetic field. Using only a few transistors these space 'bots seek the brightest available source - the sun - and orient themselves precisely toward it. Acting like a "delay line", the researchers say, electronic pulses from photodetectors travel first to one neuron and then the next. By adjusting the timing relationship between the pulses, the control system seeks the light and uses the reaction of the controller's magnetic field against earth's magnetic field for the torque needed to orient the satellite. With six neurons on three axes, the controller can move or examine different points in three-dimensional space.

Since robots and spacecraft are immersed in their environment, environmental feedback into these systems is unavoidable. Whenever possible in conventional systems, such spurious feedback is regulated or removed. However, in Nv systems, say the researchers, the main advantage of their adaptive quality comes from the fact that the actuators themselves are effective environmental sensors. For example, an Nv-equipped satellite that is oriented using magnetic torque coils will respond to the magnetic variations induced by geomagnetic storms, even if it has no magnetometer for sensing the ambient field. Thus, the sea change in satellite design: instead of designing systems that bully or oppose their environment, the researchers are attempting to engineer systems that depend on relaxing and "flowing" within the environment.

Tilden says hundreds of microsatellites in orbit could relay simple streams of data to a communications mothership, which could integrate the data and pass it to ground stations. Measuring earth's magnetosphere is one potential mission. By locating a swarm of microsatellites on the sunward side of the magnetopause, scientists can take real-time measurements of the energy transferred to earth's magnetic field by the turbulent solar wind, something we've never been able to monitor.

These three-inch satellites would be small enough to orient themselves within a single fluctuation in space plasma, says Moore. "These microsatellites can go where expensive, big satellites can't go and they can perform a class of business and science missions that no other platform can. Nobody knows how small you can go. That's what we aim to find out."

When might one fly? Technically, the researchers still must micro-gravity test one of the prototypes and then fit its control attributes into the standard aerospace engineering conventions for autonomous spacecraft. Funding has become an issue.

The research, which once received in-house support from a Los Alamos lab called Dynamically Adaptive Processing Systems, is currently unsupported. "In the meantime, progress is made on our own time and nickel," says Tilden, whose lists among his heroes visually scientific meme designers Syd Mead, Osoni Umetsu, Wayne Douglas Barlow. "Right now it's science only. Personally, I think the goal is to populate space and planets with so many of these devices that they serve as an artificial robo-ecology, bringing manned missions down from the arduous, ultimate camping trips they are now."

Tilden says the problem with his research is that it's just research. "Where it's headed depends on how many I can convince that computers are not good for everything and that there are other technical solutions to interstellar exploration and colonizations on the cheap."

He's negotiating with a LANL department now that has an aging satellite in orbit. "We're asking that before shutdown they load a simulation of one of our satbot controllers into it. That could happen soon."

It could also not happen at all, in which case Tilden isn't disposed to worry much while these creations await their chance to seek the light. "The satbot project is just one of many right now, so I'm letting it run on automatic just to see what kind of interest it generates. It's such a violation of conventional aerospace paradigms however that it will likely get re-shelved. That's okay. My prototypes have expected lifetimes of decades and they're happy enough hopping about on my window ledge awaiting their chance."

In the meantime, Tilden contents himself dreaming of founding an institute to explore all scientific aspects and applications of autonomous robotic devices. "Now that I'm convinced my Nv controllers make everything possible, I find there is insufficient time to explore all possibilities. The institute would be stocked with BEAM types, and we'd tackle everything from giant walkers to nanoprobes, materials to actuators, robocrystalography to self-reproductions - but primarily, how robots might serve Planet Earth best."


Comments? Questions? Corrections? Assignments? douglaspage@earthlink.net
Return to Home Page.
Comments? Corrections?  Questions? Assignments? douglaspage@earthlink.net
Back to Top. Return to Home Page.