FAQ
From BEAM Robotics Wiki
Answers to Frequently Asked Questions
This FAQ is intended to serve as an immediate resource for those interested in BEAM robotics. If you have a question not answered in this FAQ, please ask it on the BEAM mailing list.
The BEAM FAQ has a long and storied history, of which this Wiki is the most recent incarnation. Mark Dalton created the original FAQ in 1996. Brian Bush picked up the torch and maintained the FAQ until 1999. Solarbotics has preserved Bush’s site, known as the BEAM Tek FAQ, at faq.solarbotics.net. The BEAM community moved the FAQ to this Wiki in 2006.
[edit] What is BEAM robotics?
BEAM is a acronym standing for: Biology, Electronics, Aesthetics and Mechanics
The basic principles of BEAM rest on the fact to build smart machines one must first build a smart body. Forget the brain, let us just focus on a simple stimulus-response based ability within a machine. In it's shortest form is:
- Use minimalist electronics (keep it small & simple, reduced parts count)
- Recycle & Reuse components out of technoscrap
- Use solar power, if possible
More elaborately, from the father of BEAM robotics:
...The science behind the idea stems from current concepts in artificial intelligence (AI), artificial life (ALife), evolutionary biology, and genetic algorithms. It seems that 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 problem is that such a concept requires self-reproducing robots which won't be possible to build (if at all) for years to come. A solution, however, is to view a human being as a robot's way of making another robot, to have an annual venue where experimenters can let their creations interact in real situations, and then watch as machine evolution occurs. In other words, robogenetics through robobiologics.
[edit] What kind of technology is in these machines?
BEAM robots use simple analog circuitry. The two most common technologies are Solar Engines and Nervous Networks (Nv Nets).
An excerpt from "Biologically Plausible Dynamic Artificial Neural Networks Reviewed" by Mark W. Tilden:
Simple, yet elegant solutions. A variety of robotic devices with adaptive nervous systems not exceeding the equivalent of 10 neurons. These devices not only learn to walk from first principles but can also adapt to many different circumstances including severe personal damage. There are no processors involved; my most complex device uses only 50 transistors for its entire spectrum of behavior, response and control.
[edit] What is a BEAM Solar Engine?
The simplest BEAM robot can be constructed around a two-transistor circuit, called a Solar Engine or relaxation oscillator. Such a BEAM robot is powered by light and twitches with movement.
The purpose of a solar engine is to act like a power "savings account" -- a small trickle of incoming energy is saved up until a useable amount is stored. This stored energy is then released in a burst, in order to drive some useful (if only sporadic and incremental) work.
The [BEAM Wiki] and the BEAM Reference Library has a collection of [Solar Engine] schematics.
[edit] There are so many different Solar Engines, which is the best?
The answer depends largely upon the robot it will power. At the time of this writing, the Miller Solar Engine is the best in terms of power output to part component.
[edit] What is a BEAM Neuron? What is a BEAM Nervous Network?
Many BEAM robots feature artificial nervous systems made up of BEAM neurons. A BEAM neuron is a resistor-capacitor network and a logic gate. Generally the logic gate is an inverter or buffer, but other gates may be used.
A number of artificial neurons can be combined into BEAM Nervous Networks or Nv Nets. This forms a non-linear analog controller that can be used for rovers and walking machines.
An excerpt from "Biomorphic Robots and Nervous Net Research: A New Machine Control Paradigm" by Mark W. Tilden
Using Nv nets, highly successful legged robot mechanisms have been demonstrated which can negotiate terrains of inordinate difficulty for wheeled or tracked machines. That non-linear systems can provide this degree of control is not so surprising as the part counts for successful [Nv]] designs. A fully adept insect-walker, for example, can be fully controlled and operated with as little as 12 standard [transistor]] elements.
[edit] What is an Nv Topology? What is a Chain or a Core?
A BEAM Topology is a way of connecting BEAM neurons into a single unit. Note that a robot's Nervous Net may contain multiple topologies. Common topologies include the chain (or string) and core (or ring).
We generally use a numeric prefix (bi-, quad-,etc.) to denote the number of Nv neurons. Then, we state the topology in the root. For example, there are Bichains, Trichains, or Quadchains for strings of two, three, or four Nvs. Similarly cores, where we have Bicores, Tricores, and Quadcores.
An alternative naming convention for topologies is the number, the neuron, and the topology. Hence, 8-Nv core or 2-Nu ring. This is particularly important when it is not a homogenous Nv topology.
[edit] What is an Nv Architecture?
A BEAM Architecture is a means of connecting various Nervous Net topologies into a complete control system. There is no hard and fast naming convention here. The most commonly used is the Master/Slave Bicore architecture. While a Nervous Net may contain multiple architectures, it is more common to have only one.
[edit] What is a Microcore?
The Microcore combines the Quadcore (4-Nv) topology with accompanying circuitry. The Quadcore uses 4/6 of the 74xx14, leaving two inverters left over for additional neurons. This is the most common configuration.
The VBug robots used the Microcore architecture.
An excerpt from "Theoretical Foundations for Nervous Networks and the Design of Living Machines" by Mark W. Tilden:
... The smallest possible Nervous network (defined as the "Microcore") for a capable quadruped with 1.25 degrees-of-freedom per leg. It features twelve transistors in a single hex-inverter chip, eight used for the Nv Microcore, and four more for a primitive neural-net "brain" so that the machine can sort itself out on power-on, and reverse when encountering an obstacle.
Though many other complex Nv architectures were tried, this was the first such design that shared up to three motors from the same neuron in a tight, adaptive structure.
[edit] What is a Unicore?
The Unicore is an architecture comprised of Bicores and supplementary neurons. The preferred IC is the 74*240, which has eight inverters. All eight can be used to implement neurons, as in the BEAMant. Another option, seen in Strider, is to build a Bicore and use the remaining six inverters for a buffered motor driver.
Robots built on the UniBug model feature the Unicore architecture.
[edit] What are the Hextile boards?
Hextiles were six-sided, gold-on-black boards that came pre-arranged in sheets to provide a convenient collection of components for BEAM robots. Unfotunately they are no longer available.
For a more comple descriptin see: Hextiles
[edit] Are there available alternatives to Hextile boards?
Solarbotics at one time sold the very useful Bicore Experimenters PCBs (BEP), but they have been out of stock for many months. The BEP boards are square in shape and are clearly inspired by the Hextiles. The BEPs can be used to build several BEAM circuits, Nv Nets, and complete robots.
[edit] Is there a mailing list?
BEAM has a very active discussion group on Yahoo groups. There over a thousand members who discuss BEAM related topics thru hundreds of emails each month. You can subscribe by going >here<
[edit] Upcoming Events and Competitions
A list of current events is maintained on this wiki.
[edit] Are there any BEAM Robotics groups?
A list of BEAM Groups is maintained on this Wiki.
[edit] What publications are available that discuss BEAM robotics?
There are several papers and books that cover BEAM and BEAM-related topics. The Living Machines paper, by Brosl Hasslacher and Mark W. Tilden, is an excellent place to start in on the theory. Check out Junkbots, Bugbots, and Bots on Wheels by Dave Hrynkiw and Mark W. Tilden for application.
The BEAM Reference Library' has many more papers available for download. There is also a recommended reading list' on this wiki.
[edit] What are the skills required to build BEAM machines?
Electronics Basic electronics knowledge Soldering skills Probably the best book for learning electronics is "Forrest Mimms" book "Getting Started in Electronics" available from Radio Shack (front cover, catalog number 276-5003)
Mechanical skills Construction skills Solarbotics - Pictures of what others have built to spark your imagination. Miller's Leg Mechanics and, of course patience.
[edit] Where can I find electronic and mechanical parts?
Many of the parts for BEAM robots can be recycled from broken consumer electronics. Techno-scrap ideal for salvaging includes anything with motors, such as floppy disk drives, CD/DVD drives, printers, et cetera.
The BEAM Reference Library contains tips on salvaging parts as well as a list of BEAM vendors.
The BEAM Wiki also has an open list of BEAM Suppliers
[edit] A Course of Study in BEAM Robotics?
A good course of study to those just beginning in BEAM robotics might be:
[all along you should be collecting and searching for tek junk]
(1) build one (or both) from the SunEater series with SunEater II first, followed by SunEater I. Please follow the specifications
(2) build a solar roller, or some solar engine-based gizmo
(3) build a photovore - study phototropisms involved and their interesting behaviors,
(4) read Tilden's patent, Living Machines and other on-line documents
(5) construct a microcore circuit (a four-neuron nervous network) and then
(6) onto the construction of a microcore walker.
(7) construct a BiCore
(8) study the phototropisms involved in nervous networks, such as the suspended BiCore
(9) start hooking up nervous networks into patterns found in nature (such as Cockroach chewing controllers)
Of course, this is only my opinion, but it begins with the simple and moves to increasingly complex designs. In all, start simple and work your way up. Since practice makes (*almost*) perfect, build, build and build. All along you should continuously search for parts and junk. I have always found a good way to build neat devices is take mechanisms you find in nature and in high-tek and try to replace them with nervous networks and solar engines. Example: a stepper motor usually has a computer as a controller. Why? Well because of the somewhat complicated patterns needed to drive the motor it is usually easier produced by a computer (program). The action of controlling a stepper motor can be performed by loops of nervous networks. Of course, this is not the most efficient manner, but it can be done. You will be amazed at the type of behavior you can get from simple loops of circuits, and it doesn't always have to be formed or embedded as a robot.
Always buy sets of things, so you can experiment and sometimes unfortunately destroy parts. Note: I am assuming you know basic electronics (see Forrest Mimms' Basic Electronics) and how to solder.
The following is a simple procedure that can be applied to most endeavors:
do simple things first
learn to do them flawlessly
add new layers
don't change the simple things
make the new layer work as flawlessly as the previous layer
repeat
[edit] What is the next step in the evolution of BEAM robotics?
Here are some ideas off of the BEAM mailing list and from Tilden himself: Vision. Mark Tilden has been doing some neat work with foviation, building light-seeking vision/head mechanisms powered by Nv technology.
Hygene. If you consider the problems of an autonomous machine with an extreme lifetime, the big question is how to keep yourself clean so that your performance remains constant. The sleek look and grooming habits of insects is to optimise this characteristic. We can make machines robust against such things (teflon wire insulation on many surfaces), but what does this mean when you've got dirt splattered on your sensors? Make a machine that can negotiate mud in the rain and still keep track of itself, and you're well on the way to an advanced life form.
More recently Tilden has been doing some advanced work into the nature of minimal cognitive architectures. Better brains for BEAM bodies involving Nv arrays, some with hundreds of neurons. The resulting robots will be... big.
