Inflatable Robots by Otherlab: A Walking Robot (named Ant-Roach) and a Complete Arm (Plus Hand)

Inflatable Robot

I'm really excited about inflatable robots... they have the potential to be low-cost, lightweight, extremely powerful, and yet "human safe" -- ie. perfect for many robotics applications.  With that in mind, I would like to introduce you to two new (breakout) inflatable robots: a 15-foot-long walking robot (a Pneubot named Ant-Roach) and a complete, inflatable robot arm (plus hand).  Both of these robots were developed by Otherlab as part of their "pneubotics" project (in collaboration with Meka Robotics and Manu Prakash at Stanford University), with some funding from DARPA's Maximum Mobility and Manipulation (M3) program.    These robots use textile-based, inflatable actuators that contract upon inflation into specially-designed shapes to effect motion.   Since these robots are built out of lightweight fabric-and-air structural members and powered via pneumatics or hydraulics, they exhibit large strength-to-weight ratios.  For example, Ant-Roach is less than 70 lbs and can probably support up to 1000 lbs; the inflatable robot arm is less than 2 lbs and can lift a few hundred pounds at 50-60 psi.  Be sure to read on for details and lots of videos!


A Pneubot Named "Ant-Roach" : The Inflatable Anteater-Cockroach Robot


Ant-Roach the inflatable robot  Ant-Roach Inflatable Robot Elephant

To quote Otherlab's recent blog post:

Here is the Otherlab’s 15 foot inflatable walking robot, the Ant-Roach.  We thought this conceptual elephant looked more like a cross between an anteater and a cockroach.  The goal of building the Ant-Roach was to demonstrate the carrying capacity and high strength-to-weight ratios possible with inflatable structures.


Ant-Roach relies on a number of fabric, inflatable actuators (left)  and pneumatic piping (middle) to move.  During my conversation with Saul, he told me that Ant-Roach weighs in at less than 70 lbs -- making it human-transportable (right) -- and yet it can probably support up to 1000 lbs (making a couple-person payload no problem).

Inflatable Robot Actuators (Ant-Roach)   Inflatable Robot (Ant-Roach)  Underbelly  Inflatable Robot (Ant-Roach) Carried


Videos of the robot in operation: 

There are a lot of additional, supplementary videos here (below).


With a little refinement, I could imagine several applications for these actuators beyond just robotics.  Heck, you could just augment those huge, inflatable water toys with actuation.

Inflatable Water Toy  Inflatable Water Toy (The Blob) Inflatable Water Toy (Giant Pyramid)


Alternatively, just turn the Pneubot Elephant into an amusement park ride:


Note to my (now-former) labmates: You guys need to hurry up and get married if you want me to bring a live elephant to your wedding procession in India.  If not... I might just bring one of these instead!



Lightweight, Low-Cost Inflatable Robotic Arm


Inflatable Robot Arm and Hand  Inflatable Robot Arm and Hand  Inflatable Robot Arm and Hand

Inflatable Robot Arm and Hand   Inflatable Robot Arm and Hand

To quote from Otherlab's blog post:

Otherlab’s “pneubotics” program has produced the world’s first inflatable, robotic arm.  We’re refining fabrication techniques and design, but this prototype demonstrates the potential strength and dexterity of this low-cost, safe robotics technology.


According to my conversations with Saul, the arm alone (sans valves and air supply) weighs a mere 2 lbs and is still able to lift several hundred pounds (eg. a person) with just 50-60 psi.  He tells me that it can handily defeat a human at arm wrestling.


Much like Ant-Roach, the inflatable arm is made from a series of fabric "pockets" that expand and contract when inflated to create motion.  The same applies for the hand:

Inflatable Actuator  Inflatable Actuators

Inflatable Robot Hand  Inflatable Robot Hand  Inflatable Robot Hand





Inflatable Robotics: Discussion


Inflatable robots have many desirable properties.  Because they are constructed from fabric and basic pneumatic parts (eg. an air supply, valves, and tubing), they can potentially be very low-cost.  Furthermore, they possess high strength-to-weight ratios owing to the strength of pneumatics and the mostly air-and-fabric structural members.  Plus, they are naturally compliant, which makes them (at least somewhat) human-safe.  Of course, there are drawbacks... 

  • I'm sure that controlling the end effector (position / velocity / force) remains a challenge.  There are no joints (and hence no joint encoders) for forward kinematics, and there is currently no built-in force sensing.  I'm sure Saul would welcome a way to perceive the robot / actuator state.  Perhaps computer vision or depth camera (eg. Kinect) solutions would be the most approachable way to track the robot's position.  As for force sensing, I could imagine wirelessly-powered force sensors at the end effector (ooh, that could be a fun collaboration for me!) or some form of robot skin.
  • Also, the on-off nature of valves is inherently non-linear.  I imagine this introduces some tricky controls problems for actuator sequencing (and coupling between the various degrees of freedom).  This will likely be an open (and rich) area of future exploration. 


Otherlab seems like a natural birthplace for inflatable robotics.   Otherlab (co)founder and MacArther Award Winner, Saul Griffith, has a long history of developing low-cost "smart material" solutions in various problem domains: low-cost lenses using the boundary conditions of an inflatable membrane (left), high-strength fabrics for kite surfing and kite-based power (a la Makani Power), and more recently (at Otherlab) "computational manufacturing" to build complex 3D shapes (middle and right).

Low-Cost Lens Manufacturing Technique Inflatable Animals from Otherlab Elephant made from metal tiles


Incidentally, Otherlab is a "private R&D company" -- a lean, think-tank-like organization with a healthy combination of hard science, engineering, creativity, and entrepreneurship.  They are a canonical example of the business (model) I someday hope to co-found (or at least be a part of).


For some related work...  I recall reading about a soft, inflatable robot by Chris Atkeson (PI) and crew at the CMU Quality of Life Technology Center. It seems they are also interested in inflatable robots for safe human-robot interaction.   From the CNET photo-op:

Soft Robots are a project from the Quality of Life Technology Center designed to help seniors and people with mobility problems.

Soft, inflatable arms like this one can manipulate objects weighing up to 500 grams (1.1 pounds) and assist with tasks like feeding, dressing, and transfers from bed to wheelchair. 

Inflatable Robot Arm Cleaning  Inflatable Robot Arm  Inflatable Robot Arm Cleaning

It appears that the lead student on their project, Siddharth Sanan, recently interned at Otherlab.




Supplementary Videos




When I saw this stuff I thought it was one of the coolest innovations in actuator and robot design.  I think like Travis said, one of the difficult problems that if solved could increase the application domain for these guys is being able to do state estimation and sensing in general.  Even until that happens, this work is really cool.

—Marc Killpack

I love the idea of using this concept for very lightweight Mars exploration hardware. They're light for their size, they can expand out of a smaller container, and they have no bearings or joints (so impervious to sand/dust).

Furthermore, their comparatively large size and high strength/weight would make many difficult tasks (moving boulders, digging trenches) very easy. These lightweight robots would make excellent companions for human explorers, since their size and strength would nicely complement the precision and dexterity of the human body.

In fact, with the right tools and pre-positioned supplies, a few of these ant-robots could probably build a human-scale habitat and have it ready to go when the humans arrive!


Wow, when I saw the first pic I thought of the Tachikoma from "Ghost in the Shell."

Why can't you just put the hoses INSIDE? Seems obvious.


Wow, this article received quite a bit of attention across the web: Slashdot, Popular Science, Time Magazine, io9, Gizmodo, Engadget, BoingBoing, IEEE Spectrum,, etc.

I kinda expected this... The work is mad-cool and has some fun videos, but I'm not sure that the general public appreciates the potential impact.  This is reflected in the quality of reader comments across the various news sites: some comments are well-founded and insightful, but the vast majority are abysmal.  Naturally, this is not the case for Hizook readers; y'all are too informed for that.  ;-)

Anyway, I think I might do a follow-up in this Hizook comment thread in a week or so (after the
articles have had some time to simmer) to pick at various reader comments from across the web... that should be fun!

—Travis Deyle

Bummer that Flash is used for the video and not HTML5 - can't watch on a tablet :(


The walking cycle looks pretty awkward and slow. I wonder if creating a pneumatic inchworm might not be easier and simpler. Maybe a catamaran inchworm which can turn by having one side crawl faster than the other.

Tablets can be used to watch flash based videos.


You're right... lots of folks (eg. NewScientist) have talked about using inflatable robots (like an inflatable Rotundus) for Mars exploration due to their lightweight properties.  Having manipulation capabilities just raises the bar...

—Travis Deyle

Core77, a really cool design website that I follow, posted some interviews (parts 1, 2, and 3) that featured interviews with Saul about Pneubotics and the invention process. They're worth watching:


—Travis Deyle

What a great idea.  Could make all sorts of fun and far out creatures........or ideas!