Helio-Courier

Joshua Tyler, HP Labs

(Note: There is video of the blimp in action!  See below, under "Maiden Voyage")

The Dream

If you, like me, are tired of having to manually deliver documents or other items within your office building, and if your building has high ceilings, good lighting, and minimal air currents, then you will inevitably reach the same conclusion I have: An automatic helium blimp delivery service.

The Challenge

Helium is lighter than air.  But not that much lighter than air.  A modest size blimp such as mine (roughly 75 cm in diameter) provides only a few ounces of lift--barely enough for a small motor and battery.  Therefore, I wanted to avoid adding any additional sensing or navigational equipment to the blimp itself.  Instead, I augmented the environment of the office to track and control the blimp.

The other main challenge in this project was a lack of time and resources.  I have a real job at HP Labs, and no part of it involves designing and testing blimps.  If you'd like to see some of my "real" work, see my home page or my lab's home page.  As a result, most of this work was done at night or on weekends, which was also necessary because bright red-and-blue buzzing blimps tend to be distracting to cubicle dwellers.  What's more, I have virtually zero experience with computer vision, aerodynamics, or inert gases.  All I have is a dream.  Ok, enough disclaimer.

The Equipment

The blimp itself is a hobby radio-controlled (RC) blimp from Plantraco, and comes with an RC transmitter than can be connected via USB to a PC and controlled by connecting to a TCP socket. 

The cameras are Logitech QuickCam Pro 4000s, purchased at a local electronics store.

For video capture and processing I used the Microsoft DirectShow SDK and Visual C++.

The Approach

I work in a relatively large, open, mostly gray-colored building with high ceilings and full of cubicles (typical Silicon Valley office environment).  I mounted a couple of cameras on top of my cubicle to survey a portion of my floor.

I intended to make the blimp stand out from the background environment by giving it high RGB values.   I painted the blimp half red and half blue so that I could estimate its orientation by measuring how much of each color was visible.  In the picture below left, the detected red and blue portions of the balloon are highlighted.  On the right is what the balloon looks like before any image processing is done (note: the photos are NOT from the same moment... that's why the orientation is different in the two images).

Using DirectShow, I wrote a little program that would:

• capture the incoming video stream
• find the blimp in the picture
• determine its position and orientation
• decide what command should be issued (e.g., up, down, forward full, rotate, etc.)
• connect to the blimp transmitter and issue the command

I had originally planned to use two or more cameras to triangulate the blimp's position.  I found, however, that one camera was pretty effective--all I had to do was measure the blimp's size on the screen to learn how far away it was.  Also, I was having a lot of problems capturing multiple streams simultaneously.

In the images below, the video stream is overlaid with the distance from the camera and other tracking information.

		The control loop for the blimp is as follows:  Climb until in sight of the cameras, then orient itself towards the target (by rotating), fire both fans forward until the target distance is reached, then descend into the target cubicle.  In practice, it is not nearly so clean, and there is a lot of adjusting of orientation and height before any forward progress is made toward the target.

Maiden Voyage

The time eventually came when I felt ready to try delivering something.  I had originally hoped to deliver documents (i.e., sheets of paper) but unfortunately found that one sheet of paper is too heavy for the blimp to lift reliably (I was serious when I said the helium doesn't provide much lift).  I settled for a Post-it Note.

In this demonstration, I am sending a message to my friend and manager Bernardo Huberman.  First, I summon the blimp and attach a note to it (shown in the image at the top of this page).  Next, I tell it to go to Bernardo's cubicle (not very far away).  Finally, the blimp descends into his cube, where he gives it a friendly greeting.

VIDEO OF THE BLIMP:
(Real, 3.4MB - best quality)
(MPG, 3.0 MB)
(AVI, 9MB)

Lessons

This project turned out to be very difficult, especially for someone with my limitations and constraints.  I have achieved the creation of a blimp that can deliver very small notes to people within about 40m of my office.  But it is a proof of concept, anyway, and some of my problems are probably only a matter of engineering and resources.  For example, I need a bigger blimp, so that I can lift more and increase stability.

HP Labs does not seem (at least to me) to be a windy environment.  An extremely light blimp, however, is very sensitive to even the slightest air currents.  Fighting these currents was, I would say, the toughest challenge I faced in this project.  At times, the breezes, imperceptible to a human hand, were sufficient to completely overpower the blimp's fan motors.  A larger, less round blimp seems like an obvious thing to try.  My next blimp will be oblong.

With my circular blimp, the problem of orientation was also significant.  The slightest movement or breeze could set the blimp spinning, and it is difficult to stabilize the blimp once this happens.  Perhaps some fins or something will help.

Other Work on Blimps

There are some other (more legitimate) research projects on helium airships of various kinds.  Eric Paulos and John Canny at Berkeley have a list of blimp-related links.  It's probably as good a place to start as any.

Contact

Feel free to contact me about this or anything else.  I'm a friendly guy.

My home page at HP Labs.

My email: joshua.tyler@hp.com