"BEAM Me Up"

Page 1 of 2

"I think Dr. Seuss might be onto something here."

I think I finally understand how BEAM robots work. How can something so simple be so confusing? That's what we will look at this week. You may have heard of BEAM robots with the recent press they have received with articles in Smithsonian and Circuit Cellar (see the BEAM stiquito). BEAM robots are the brainchild of Mark Tilden. BEAM is a triple acronym that stands for:

I stumbled onto Mark Tilden's work for the first time last winter while surfing the net for inexpensive robot plans. My 6-year-old son wanted to build a robot that would clean his room. I thought I might nurture this creative bent, so we started looking around for some ideas. I have to be honest, the first time I took a look at one of these designs, I just couldn't get it. "How could this thing work consistently?" I thought to myself. I couldn't really grasp the concept well. So we ended up with a Lego Mindstorms robotics set instead. We have had a lot of fun with the Legos, but that is another story. For the most part, I basically forgot about Tilden's critters and went on with life. A few months later, my third son was born. Being an engineer at heart, I would watch him develop and wonder at the way he learned to grasp objects and recognize voices. I noticed he spent a lot of time simply flailing his arms and legs wildly in a random fashion that would occasionally sync up to something useful. During these weeks of reflection on the nature of intelligence, a colleague at work brought me an article on "robotics" from the Smithsonian. I had told him about BEAM robots back when I first looked into them, and he thought I would be interested in this article. It was titled Redefining Robots, and I highly recommend reading it.

What is a BEAM Robot?
I would describe it as a robot with a relatively simple circuit that can solve unforeseen problems in its path. Much like an insect. Insects don't have a lot of brainpower, but they can negotiate the world better than many high-tech, tera-flop loaded, processor-based robots currently in development and use.

The first robot I read about that Tilden made was called the Walkman. It is a 4-legged critter with a motor for each leg that would pulse in sequence to produce a walking motion, but the cool thing is how it would negotiate terrain and obstacles. It would seem to try various approaches until it found one that worked. Oh and by the way, it doesn't have a microprocessor in it.

The Basic Design of a BEAM Robot
We will go through the heart of a 4-legged walker robot to get a feel for how these work. First, take a look at this diagram, it is called a "nervous net" node.

Figure 1

It is pretty straightforward; you should be able to see that a step input will create a pulse output based on the RC time constant. With this connected to a motor, the output will cause the motor to run for a while. Simple really, but the magic occurs when you do this:

Figure 2

What will tend to happen here is a pulse will go around this circuit repeatedly in sequential order, pulsing each motor so that the beastie will walk. (You have to watch out for more than one pulse chasing around the loop though, sorta causes the robot to have epileptic seizures.)

One very important thing to understand: The connection to the motor must be a two-way street. Motor loading must affect the input to the next node. That is how you get the cool behavior. To really spice things up, start adding sensors that disrupt the various nodes. For example, a feeler might make the pulses on one side last longer so as to get the unit to turn.

So What Does a BEAM Robot Do?
Well that's a good question; right now these machines basically exist and survive. But they are starting to develop into more complex designs. At Los Alamos, they are interfacing one nervous net into another, so you have a type of stacked or layered network system, and they are seeing more complex behaviors develop as a result.

There are useful possibilities because of their low cost, such as wandering around a minefield and blowing themselves up, for example. (The pyromaniac in me would love to watch that.) In many ways these robo-insects are still a curiosity. But who knows, we will just have to wait and see.

How Does It Work?
Like I said earlier, a lot of what happens here is in the feedback. But Tilden says the circuit is chaotic in nature, and we don't really understand chaos well at this point. An interesting thing is that the circuit does try to minimize the energy it has to expend at any given point. While I was explaining this to a co-worker, he told me that back in the 1600s some guy discovered that if he put two pendulums on the same shelf and started them swinging in opposite directions, eventually they would sync up because that is the minimal energy state. If any of you know more about this bit of history, please e-mail me—there might be a clue here. To summarize, no one is really sure, and so far no one has be able to model chaos on a computer either.

Next >>

Product Engineering Archive

		Click here to get your listing up.