Figure 2

(Bypassed)

(Input)

(Output)

(Input)

(Output)

A BEAM Technology

To envision life on a pile of metal, silicon, wires, gizmos, and widgets has been the dream of many people — from the ancient Greek writers' legends of giant metal men to George Lucas' sci-fi saga. Thanks to current pioneers — such as Mark Tilden and Rodney Brooks (among others) — we are getting closer to achieving this lofty goal.

In the last few years, we have seen artificial life, chaos theory, and BEAM Robotics1 actually become legitimate sciences. In the new field of Artificial Life (known as A-Life), biological behaviors — such as emergence, evolution, and flocking — appear in non-living entities2. We have seen the classical notion of artificial intelligence give way to neural networks, fuzzy logic, and emotional robot heads, like Kizmet3. Nowhere is this exciting evolution more raw and passionate than in the science of BEAM robotics, but — like most robotics fields — BEAM is in its infancy. In BEAM's present state, there are limitations — or are there?

Most BEAM roboticists study a creature and its effects on its environment using a minimalist analog approach. In fact, attempts have already been made to establish rules of behavior, such as Mark Tilden's Laws of Robotics, found in Junkbots, Bugbots, & Bots on Wheels:

1. A robot must protect its existence at all costs.

2. A robot must obtain and maintain access to a power source.

3. A robot must continually search for better power sources.

by Jason Travis

Startling work pertaining to emergent behavior has shown success, even though these creatures are still fairly basic in the behaviors they exhibit. It is impressive that a robot — such as Mark Tilden's Walkman — can navigate over the desert terrain around Santa Fe, NM. This is a task that robots from universities all over the world find difficult — even impossible — without huge amounts of computing overhead. Walkman does it with a couple of ICs, resistors, and basic circuits, all used in unconventional combinations. When faced with an obstacle, the Walkman either powers over it by warping its leg and body or backs up and heads in a new direction.

All this has happened while the traditional university mobot is still determining the type of terrain and calculating the best course. However, while BEAM bots excel at crossing terrain, they cannot report back to base what they see, as the Mars Rovers do. BEAM bots are good at instinct, but they are not that good at thinking; they are akin to insects and bacteria in this regard.

The point of all this is to stimulate and prepare the robotics community for Phase II of the BEAM evolution: the coming of BEAM ecology. We are going to build an ecosystem in our minds. Maybe one or more of us might choose to extend this into the real world and build a living ecosystem. This would, of course, entail mostly the use of the biology aspect of BEAM. What I am proposing is a fusing of the world of BEAM and the world of A-Life. As the lines blur here, I will leave it to the reader to differentiate between real and artificial, living and non-living. Beware — beyond these edges, there be dragons.

Earlier, I mentioned Tilden's Law Number One: Protect thyself. What if, in this ecosystem, some of our BEAM bots could protect themselves? How about hard-wiring in the concept of live, don't die? Put into biological terms, this is known as fight or flight. Any student of BEAM technology has heard the one about the cat getting hold of a legged BEAM Walker, mangling it, and, yet, it still walks. This is a testament to the robust nature of such a bot. What about arming a BEAM robot? We already do it in robot wars, games, and even sumo. How about a BEAM robot with a stinger — say a whipping tail with an electromagnetic disrupter? How about a robot with a strong jaw and saw-like teeth to bite with in a confrontation?

Let's take an evolutionary view of our ecosystem. If a bot's batteries die and it cannot get to a recharger, do we take it out of the world? Should we leave it there to decompose? What if the bot's solar cell becomes disconnected; do we repair it? I suggest the answer is leave it.

After all, when we die in this world, we aren't reanimated any more than a deer is when it dies. These are ideas that are unique to a closed ecosystem: ideas such as old age, other competitive creatures, predators, natural obstacles, diseases, and even starvation.

How about reproduction? Okay, I know what you're thinking — we can't even get a robot to build a copy of itself, let alone reproduce (with the exception being a couple of Japanese automated factories that build robots, but they still require human interaction). Therefore — as is sometimes necessary — we will make an exception to the

References definition of reproduction. We will use external introduction to substitute for reproduction.

There are many examples in nature of non-sexual reproduction. Asexual reproduction is a good example, as is the invasion of non-indigenous species into a fixed ecosystem, causing crossbreeding. What if we add a new BEAM creation into our ecosystem every month? Now a "new" bot is born from the perspective of that world.

How would it affect the existing bots? What about surrogate mothers? What if we produce a weird bit of hybrid behavior between two robots? One robot is the surrogate mother who waits for the "baby" to arrive. When you introduce the baby into the new world, it doesn't have any power. The mother bot has to pick up the baby bot and carry it to a power station to feed it — kind of like nursing.

In nature, there are a few creatures that even use community nursing. Another take could be that the baby has a limited amount of power when born and has to cannibalize power from the mother bot, who willingly

1 BEAM - Stands for Biology, Electronics, Aesthetics, Mechanics. Biology means to make it comparable to nature in function and design. Electronics is short for running with basic analog and digital circuits to imitate animal reflexes, rather than using computer brains. Aesthetics, in essence, means make it pretty, not just functional. Mechanics means to make it functional; do more with less, so that emergent gives up her life for the newborn. In nature, some insects also eat their birth mothers' bodies for nourishment.

This brings us to the concept of death. BEAM robots do have a limited age; motors wear out, electronics short out, and legs mangle to the point of no return. If a robot runs out of energy and cannot regenerate it, it is dead. For example, a phototropic robot could die if trapped under an artificial plant; it runs out of power and is never again able to obtain sunlight. A recharger bot — despite being hard-wired to seek out a recharger — never makes it to one. A

behavior comes out of wire and rods.

2 See Larry Yaeger's Polyworld at http://homepage.mac.com/larryy/larryy /PolyWorld.html or the International Society of Artificial Life at www.alife.org

B See the Robot Kizmet at his MIT website: www.ai.mit.edu/projects/humanoid-robotics-group/kismet/kismet.html

predator bot can rip out the electronic guts of an herbivore bot. Alternately, a predator can eat all the energy in another bot's power pack. Imagine a BEAM predator that is designed to seek out another bot and suck the volts right out of it.

That brings us to Rule Number Two: Feed thyself. I briefly teased you with the idea of an herbivore. How about roaming herbivores that feed off plants? That brings to mind a very specialized type of BEAM robot — the BEAM plant. This would be a semi-stationary bot whose only movement would be to

New Products

(continued from Page 34)

0 0

Post a comment