Project "Gigapan"

Page no longer maintained. I no longer use Lego in my photography as I have a GigaPan unit. However, most of my panoramic images at gigapan.org taken prior to September 2008 were taken with a Lego robot of some sort.

This page is about robots I have built to help me take panoramic photgraphs. This is work inspired by Philo's cleverly-designed Panobot 2:

This is a robot (made from the old Mindstorms kit) comprising a RCX computer (the big yellow/grey brick), some levers and a cradle for holding a camera. In the Panobot the camera sits between the two black upright beams. The idea is that the computer makes the robot pan from side to side whilst pressing the camera's shutter button at regular intervals to build up a huge panoramic image. Panobot 2 is ingenious as it runs on rails(!) but is not designed for my camera and does not tilt up and down.

Robopan 1.0 - The "Spod"

Sadly no pictures of my version of Panobot survive apart from this early work-in-progress. My camera - a Sony DSC-H7 - doesn't have a straight line or flat surface on it (apart from the base but even that has a rubber foot that sticks out under the lens housing), so I found I needed to build a cage to hold the camera.

The cage and shutter mechanism

The rails

The bot didn't survive either: this picture shows what happens when something made of Lego containing half a kilo of camera and computer falls from a height of 160cm...

 

Robopan 2.0 - The "Dude"

I rethought the design and found I didn't need the complexity of the encoders. I replaced the rails with excitingly big wheels with soft, grippy rubber tyres.

To keep the robot turning around the same central point I added a vertical rod that goes through a hole in the middle of the base. Some blu-tac makes the rod surprisingly rigid.

Underside of bot showing hole for vertical centre rod

The vertical rod unit...

...and its underside

In fact the robot is sufficiently precise that it spins about its centre without the rod, but I found the rod helps to stop the bot from sliding when the base is tilted: it seems to work well even when the base is tilted at 30°.

In due course I plan to attach the base to a tripod.

The Dude in Action (6MB)

Sadly, The Dude had something unfortunate in common with its predecessor viz an inability to survive a fall from the top of my tripod loaded with the camera and the RCX:

In due course I plan to attach the base to my tripod.

Robopan 2.01

Clearly I needed to rethink my design fundamentally; get back to mathematics, physics and honest-to-god mechanical engineering:

Robopan 3.0 - The "Tacjac"

A problem that I have been working on for a long time without success is: how to make the robot tilt? Should the whole robot - including the panning mechansim - tilt, or should the tilting mechanism be mounted on the panning mechanism? Could the Lego Mindstorms puny 9V motor cope with either solution? I always felt that my robots should have a certain Heath-Robinson quality which they were lacking until I discovered the utility of string (see above). "String" suggested "pulley" which in turn lead to "winch".

The first photo shows the winch at the rear of the bot. I originally used string which had a certain je ne sais qoi, an air de la cour de ferme if you will, but in the end I abandoned it because it was naff. So the bot's winch now uses dental floss. I went for a more subdued grey and black colour scheme without realising it would be harder to take photographs of the camera. It doesn't help that I am a noob photographer.

Anyway, if you can't work it out, the camera sits in a swing that pivots about a rod. The camera and shutter motor are deliberately placed so that the swing is front-heavy. The winch stops the camera tilting forward and gravity stops the swing tilting backwards. So, to tilt up the bot winds the floss in and to tilt it lets it out. The version shown above (3.0) has no tilt-limiter so the camera will fall out (both ways) if the winch runs for too long.

The bot does have a pan-limiter. It's difficult to see in the photos above so here is a close-up:

As you can see, it comprises a lever pivoted near the bot, two touch sensors, a couple of rocks and a very intuitive, hands-on user-interface. As the camera pans from left to right the lever eventually bumps into the left-hand rock which causes the other end of the lever to press the touch sensor. The camera then starts panning from right to left until the lever hits the right-hand rock whereupon it starts scanning from left to right again.

The bot takes about 20% more images than it strictly needs to because Lego is not a precise medium for building robots. However, I was pretty pleased at the result. The bot took 210 photos while all I did was sit and watch for 20 minutes. (For "sit and watch" read "worry about the floss breaking, the camera falling out, fat sparks and blue smoke coming out of the motor, Aeriscera's Theorem etc.")

Aeriscera's Theorem

For every Lego robot R there exist points P and Q on the robot and vanishingly small forces FP and FQ such that when the forces are applied at P and Q respectively, the robot will explosively disintegrate.

Proof

Impossible to present - this page has no margin.

 

Robopan 3.1 - "Die Flahne"

The Tacjac was quite successful but the low-slung chassis was too weak for the weight of the bot. This lead to a redesign with a much chunkier base. Die Flahne, however, has something in common with earlier models: Can you guess what it is yet?

Die Flahne is a product of Fluffy Engineering - a new, radical school of thought in which it is perfectly OK for a design to fall apart in a light breeze. This philosophy leads to rapid prototyping and efficient design (i.e. no over-building). In any case, I think the idea of a robust Lego robot is against nature.

Here is a rather good picture of Die Flahne in action taken by Nic Barron:

 

The resulting pano is here.