Swarmanoids: Foot-Bots, Hand-Bots, and Eye-Bots Cooperate to Win "Best Video" at AAAI 2011

Swarmanoid System: Hand-Bots, Foot-Bots, and Eye-Bots

The Swarmanoid project is a cool twist on swarm robotics -- researchers use a heterogeneous swarm of robots to achieve distributed mobile manipulation. The swarm is comprised of three different robot varieties: Hand-Bots (manipulation and climbing), Foot-Bots (wheeled mobility and sensing), and Eye-Bots (quadrotors for recon and sensing).   The latest video of Swarmanoid retrieving a book won the "Best Video Award" at the Artificial Intelligence Conference (AAAI 2011) in San Francisco just the other day.  You can check out the robots and winning video below.



I would really like to know more about that video: assumptions, level of autonomy / scripting / teleoperation, etc.  The swarm seems too capable -- it's hard enough to get a single robot to achieve that task. In any case, it's an interesting concept and the hardware looks pretty solid too (see below). 

It seems like robot swarms (distributed systems and control) are becoming more-and-more popular -- from Georgia Tech's Khepera robots building a quadrotor landing pad, to ETH Zurich's distributed flight array (cooperative helicopter swarms).  I have a hunch that this is more of a (personal) selection bias rather than a true shift in research trends.  For example... researchers now have several low-cost and capable robot hardware platforms at their disposal (like the $300 Parrot AR.Drone quadrotor), so it is easier for them to build flashy demos that draw our attention.   Or maybe I'm way off base... let me know your thoughts in the comments.

Anyway, here are the three Swarmanoid robot variants:


Swarmanoid Hand-Bot


The Swarmanoid Hand-Bot has two grippers for manipulation and climbing, a Batman-like grappling hook with a rope that (magnetically) attaches to the ceiling to assist with climbing, and two ducted fans for rotational control when hanging from the rope. Hand-Bot is virtually immobile and sensorless; instead, it relies on several Swarmanoid Foot-Bots to move it around. 

Swarmanoid Hand-Bot  Swarmanoid Hand-Bot



Swarmanoid Foot-Bot


The Foot-Bot is responsible for ground-based sensing and mobility.  It has a small gripper that can latch onto other Foot-Bots, Hand-Bots, or objects for cooperative mobility.  For example, several Foot-Bots are required to move the Hand-Bot around.  Foot-Bot has dual cameras (including one omni) and spinning 360-degree infrared ranging sensors. 

Swarmanoid Foot-Bot  Swarmanoid Foot-Bot



Swarmanoid Eye-Bot  


The Swarmanoid Eye-Bot gives an aerial view of the environment for fast recon, and it's ceiling attachment device lets it perch for prolonged operation.  Eye-Bot is a quadrotor (robot helicopter) with pan-tilt camera, infrared distance sensors, IMU, sonar, and magnetometer.   

Swarmanoid Eye-Bot  Swarmanoid Eye-Bot



According to Swarmanoids.org, this project is a multi-national effort between several European research groups: IRIDIA (Belgium), IDSIA (Switzerland), EPFL-LSRO (Switzerland), EPFL-LIS (Switzerland), and CNR-ISTC (Italy).  You can read more about this project at Swarmanoid.org.

[ Hat tip to Lorenzo Riano. ]



Excellent projects and research  ...

Several years ago, as a graduate student, I looked into similar types of distributed robotics in a cursory manner and considered (what seems to me)   a  relatively obvious extension into a related area of research.  I never pursued the research, but instead was lead into an entirely different direction by my advisor.

Basically, I was interested in integrating the  specknets  (originated by D.K. Arvind / Edinburgh University) or smart dust particles into massively parallel systems to be used in distributed robotics.  Specks were to be protected by silicone rubber and placed in an electrolytic solution from which they would draw power --- somewhat like a wet cell battery turned inside out. With the appropriate technology, the specks theoretically could be recharged without the need for individual battery packs.

The idea was for the cells to communicate  wirelessly  using IR or optical technology.  RF technology was not considered due to the need for a prohibitively larger antenna on the individual specks. I looked into different EM waveguides for the specks, including the use of the electrolytic solution as a waveguide. This would allow communication between the distributed robotics without the need for cumbersome wires or buses.  Obviously, this  is why I initially considered the technology.

Using this technology, a complex distributed robot with parallel architecture could be built in which specks could move to any part of the robot where needed.  If  contained as individual cells within a centralized  location, the specks might function as a rudimentary brain or CNS. I should point out here, that these were distant goals well beyond the initial goal of powering specks or smart dust within an electrolytic solution.

It was at this point that I became involved with my advisor's research project, and left the above notions by the wayside. 

It is not obvious to me that either isolated specks, specknets,  or so-called smart dust is used in any of  the swarmanoid technology discussed within this article. It is probably academic since swarm robots must use some form of analogous technology in their function and communication.

I am pleased to see such promising research being explored. I hope that the swarmanoid inventors continue their research toward more complex robots and its eventual results. 

—Daniel Gray

Hey Daniel,

There seems to be a lot of research in nano (volumetrically-speaking) robots powered by electric or magnetic fields.  However, the E-or-H fields are usually used purely for actuation; often the sensing and computation is off-board.  I suppose this is mostly due to cost considerations -- custom IC fabrication is prohibitively expensive.

I'm not aware of any examples where the robots act as the second half of a wet-cell battery.  That's a  cool idea!  You wouldn't need to have a constant field applied -- the power storage is inherrant to the surrounding environment.  Someone should follow up with that... it would probably be a fun project.

A few more things...


My Favorite Field-Powered Example: Diode Propulsion


My favorite nano robot system has to be diode propulsion, which was published in Nature Materials back in 2007 (PDF: Remotely powered self-propelling particles and micropumps based on miniature diodes). 

Basically, you drop a regular 'ole diode into an electrolyte solution (saline water) and apply an alternating high-voltage electric field across the liquid.  The diode rectifies the field locally to create a DC potential.  This DC potential causes ions in the water to flow between the diode terminals due to electro-osmosis.  The reaction force due to this flow propels the diode forward.  There is a diagram and video of this process below.

Diode Propulsion for Nano Robots


As you can see in the 2nd video, you can arrange the diodes to make more complex motions.  I could imagine nano robots in place of bulk diodes, where they could multiplex the rectified voltage between various electrods to achieve arbitrary movements. 

Actually... I (briefly) tried to replicate the basic diode propulsion system early in my graduate career, but was unsuccessful.  I attribute this to my own ineptitude rather than the idea's soundess.  Perhaps it's worth revisiting.


My Work: (Macro-Scale) Wirelessly Powered Robot Swarms


One of the first articles on Hizook was about my own work on wirelessly powered robot swarms.  Basically, mini-sized robots using inductive coupling for wireless power and communication (PDF: "Surface Based Wireless Power Transmission and Bidirectional Communication for Autonomous Robot Swarms").


Wirelessly Powered Robot Swarm





I've written about to other cool nano/micro robot systems that I really like: Electropermanent Magnets and I-Swarm.  You should check out those articles too.

Electropermanent Magnet Robots  I-Swarm Micro Robots


Finally, I think the ideas of "programmable matter" and "paint-on computation" are interesting.  You might enjoy these links:

  • Claytronics Project: A CMU project to build programmable matter
  • Puspin Computing: Bill Butera (MIT Media Lab) worked on Pushpin computing, a form of painted computation.  The MIT folks developed some cool distributed applications using this system, eg. distributed postscript rendering.


OK, enough ranting.  ;-)

—Travis Deyle

Hi, just wondering where could you purchase the foot-bot robot?