From

mrc@cdr.stanford.edu

Subject

Re: Interview Request/IEEE Computer

Date

July 22, 2004 10:47:43 AM PDT

To

LDPAULSON@compuserve.com

Cc

cutkosky@stanford.edu

On Thursday, July 22, 2004, at 09:59 AM, Linda D. Paulson wrote: First, I had been asking each person to define biomimetics, but each seems to have a very different definition with some common threads running through their description.

For my group, Biomimetics is about extracting principles from biology and applying them to man-made devices -- particularly robots. We make a point that we are not trying to "copy a cockroach." This would be impractical and besides, who would want one? Instead, we've been working closely with Prof. Bob Full at the U.C. Berkeley biology dept. to learn what principles are responsible for the ability of cockroaches to run rapidly over rough terrain with essentially open-loop control.

I'd like to know specifically about your cockroach-based robot, although I understand from one source that you are also working on a bee and a robot that mimics a cat. Is this accurate? I take it the entire cockroach series of devices is named Sprawl. Correct? For what task(s) is it designed?

No. We have primarily been making cockroach-inspired cockroaches. There is a company, Iguana Robotics, that has grown out of some research by Anthony Lewis that is making a cat-inspired robot. Also Kimura in Japan has a robot, Tekken, I think the name is, that is somewhat bio-inspired. If you Google those names I think you'll find them. Those machines are primarily biomimetic in that they use a control system that is inspired by animal nervous systems. In contrast, Sprawlita & the Sprawl family draw their inspiration from the physical construction and mechanical design principles that are responsible for the robustness of the cockroach and for their ability run rapidly over rough terrain.

What was it about the cockroach that made it prime for mimickry? One other researcher with whom I spoke who is working on a different cockroach-inspired robot says they are great runners.

1. They are very fast. The American Cockroach can do something like 50 body-lengths per second. The larger tropical ones (Death's Head Cockroach) that Bob Full primarily works with are slower, but still very fast. 2. They are robust. It's hard to damage a cockroach. 3. They run over obstacles without slowing down or getting knocked off course and they do this mainly by virtue of having a wonder tuned mechanical system (sort of like the suspension of a car) that keeps them stable and on course.

How does this robot do what it does -- and please include some detail. For example, what part does computer science play in enabling its sensing or locomotion? What kind of systems -- traditional silicon chip, MEMS, VLSI, et. -- or configurations give it its abilities to move, sense or conduct other functions. Please feel free to go into some detail with specificity. Our readers enjoy and expect technical detail.

The main secrets behind Sprawlita & siblings are mechanical rather than computational. In fact, that is one of the main points. In the past, legged robots were expensive and required fast computation and accurate sensors to achieve rapid locomotion. In contrast, the Sprawl robots rely on a tuned, resonant mechanical system. The key features are:

1. 6 legs and an alternating-tripod gait in which the front and rear leg on one side are actuated at the same time as the middle leg on the opposite side to create a tripod. This is the fastest of the 'wave gaits.'

2. A sprawled posture in which the legs stick out from the body to create a larger polygon of suspport.

3. Passive compliance (inverse of stiffness) and damping at the hip joint. The swinging back and forth of the legs with each step is entirely passive. The robot operates as a resonant system.

4. Propulsion is provided mainly by the rear legs, thrusting along their axes. The front legs serve mainly to brake the robot at the end of each step. The middle legs do some of each. This same specialization is seen in insects. (Humans, with only 2 legs, require legs that are more multi-purpose.)

5. The thrusters are actuated with an open-loop motor pattern. They are driven by a clock (timer chip) associated with the on-board processor. There is no closed-loop feedback. The ability of the robot to sustain a stable bouncing trot in response to only open-loop actuation is a matter of dynamic stability, which some of our research publications address in much more detail.

6. The entire robot is constructed using a layered rapid-prototyping process that allows us to embed sensors and actuators and microprocessors in a durable polymer shell. It also allows us to locally vary the stiffness and damping to "tune" the dynamic properties. Spatially varying materials properties are ubiquitous in nature but rare in man-made devices.

In summary -- the main role of computation has been for dynamic simulation and analysis of the robots (see http://cdr.stanford.edu/biomimetics/documents/v_sprawl/main.html ) and for their fabrication via Shape Deposition Manufacturing (see http://cdr.stanford.edu/biomimetics/sdm.html )

Is this project funded by a specific government agency? How important is government funding to these types of projects in general? Are there any entities in the private sector interested in biomimetic robots for practical applications? In other words, not for toys or as novelty items. If so, which ones? Why?

The project was funded mainly by the Office of Naval Research. Government funding is essential for the kind of research that Sprawlita embodies because the applications are still a few years away. The next stage could be undertaken under applied research with a mix of government and private funding to create robots for search and rescue, for the military, etc. Good examples include iRobot ( http://www.irobot.com/home.cfm ) and the hardened version of RHex which is being produced by Boston Dynamics Inc. Note that the irobot machines do not yet incorporate the principles embodied in Sprawlita. RHex is closer -- which is not surprising as the group that produced RHex also cooperated with Prof. Bob Full.

Is this ready to be used in the field? What makes it particularly ready to be deployed? What specific types of work needs to be done on your project?

Sprawlita and iSprawl are not quite ready to be used in the field. iSprawl is closer and is almost like an RC car. But it would need to be made still more robust -- chiefly the electronics and motors would need to be upgraded and better feet would have to be fabricated. The next step, in my opinion, would be to make a version using injection-molded plastic parts and to subject it to testing by potential end users (police, fire, military).

As for toy applications -- actually they are very challenging. Radio-controlled cars are subjected to all kinds of abuse by little boys! And the companies that are making toy robots have to do extremely clever engineering to achieve entertaining performance at an acceptable cost. I think if you asked iRobot engineers about the challenges associated with Roomba versus their expensive military robots you'd find there were many.

What is the future for this field? What will the next decade in robotics look like?

I think toy applications will be important. Apparently Sony and other companies think so to. And they have far more money to invest in development than the government or the small companies that are pursuing military and rescue applications. A couple of things are driving the "new wave" in robotics:

1. low cost, low-powered computing has become powerful enough to make the robots do useful tasks in "real world" environments outside of the laboratory. Small mobile robots can now respond to visual information, sounds, touch, etc. and they can perform tasks like delivering mail in hospitals, vacuuming, giving guided tours of museums, etc.

2. Our ability to analyze, design and fabricate novel "biomimetic" devices and structures is finally getting to the point that we can really put some of the lessons we're learning from biology to practice. Ten years ago, even if I had understood exactly what materials and mechanical principles underlie the cockroach's robust dynamic locomotion, I would have been unable to build a robot that embodied them. Today, with processes like Shape Deposition Manufaturing, I can.

I appreciate your time and assistance. Please feel free to provide any additional comments on areas I might have glossed over or neglected to discuss. Also, once I've submitted a draft, please be aware that my editor will typically have questions after he's read the piece. This may require my asking you for amplification on remarks or additional questions.

That is fine. I would also appreciate a chance to see the final draft to make sure no errors have crept in.

Also, please be sure I have your correct job title and affiliation for proper attribution.

Professor Mark Cutkosky Dept. of Mechanical Engineering Stanford University

-- MarkCutkosky - 24 Aug 2004