|
Why did you build it?
Three reasons. Firstly, I needed a
project for my MEng. Secondly, I had already developed a Rubik’s cube
colour recognition system as part of a course in artificial intelligence
I had taken for my BEng. Finally, I organise and run robotics shows and
I needed a new robot to showcase.
How did you come up with the
idea?
I have always had a cube sitting on my
desk that I would fiddle with from time to time. While I was studying
artificial intelligent classifiers, I needed some data to classify for a
project. There was the cube with a nice little collection of colours to
classify so I developed a neural network based colour recognition
system. Then when I needed a project topic for my MEng, I decided to
take the previous project further to include the rest of the bits needed
to physically solve a cube. I wanted to make it look cool for the
robotics shows that I do so I designed it to look humanoid.
How long did it take to build?
I normally say 6 months but actually
this is not entirely accurate. The truth is RuBot I, the prototype took
me 5 months to build and program but I had been thinking about it for
about a year before I started. When it came to building RuBot II, all
the thinking was done and it took about a month to build and a day to
modify the program. So it actually only took me a month to build but I
had done at least 5 months research to get to that stage hence the 6
months answer.
How much did it cost?
Actually, I’ve never calculated it and
I don’t intend to. I’d be afraid of the answer I would get.
Can it beat Humans at solving
Rubik’s cubes?
Not the fastest humans. The world
record for a human to solve a Rubik’s cube at the time of writing this
is 7.08 seconds. That’s an incredible feat. Most good speedcubers can
solve a cube in an average of 20 seconds. RuBot’s average is 35
seconds.
Why is it not as fast as humans,
how about a faster computer?
The problem is not with the computer,
it is physical. Humans have ten fingers, RuBot has four. If a
speedcuber was restricted to using four fingers I am pretty sure it
would be a closer contest. When humans manipulates a cube they use all
ten fingers and can do flicks and complex multi-turns that is just not
possible with only four fingers. The human hand is so dexterous that I
think it is unlikely that we will be able to simulate all of the
intricate moves that a speedcuber can do so quickly with a robotic hand,
at least for now.
Can it go faster?
RuBot could go a little faster. I
could probably knock 10 seconds from its average time but there is a
problem with momentum. When RuBot turns the whole cube, momentum keeps
the cube turning after the hand stops moving. This is because it is
only being held by two fingers and the stop is quite sudden. RuBot’s
current speed is mainly restricted due to this. To overcome this would
require a lot more work and I think its good enough as it is and not
worth the effort for the few seconds saving.
How does it work?
RuBot picks up the cube and examines
all six sides with its colour vision. From this he works out the
current configuration of the cube. This is then run through
Kociemba’s
2 phase algorithm to find a solution with less then 20 moves. It
normally takes less than a second to calculate that. Once it has the
solution RuBot executes the moves required to solve the cube.
What is its fastest time?
Well that depends on how mixed up the cube is. If
it only has one side turned then it is 1 second. For a really good mix,
the best time I’ve seen is 26 seconds. A more interesting question is
what is its slowest time?
Ok then, what is its slowest time?
The longest I have seen RuBot (in it’s current
configuration) to solve a cube is 43 seconds. My best time is 45
seconds so he has never lost a race with me.
Why is the record 64 seconds, I’ve seen faster?
The Guinness World Records rules state that the
time taken for the robot to scan the cube is to be included in the final
time. I did not build RuBot with the intention of making or breaking
any records so this time is quite slow. It takes RuBot 8 seconds to pick
up the cube and a further 17 to scan it and find a solution. This means
that 25 seconds are included in the total time. RuBot solved the cube
in 39 seconds (a little slower than average) so that gave a total of 64
seconds for the record. I think the scan time is a bit unfair because
it is not part of the normal solving rules for humans but I can
understand why it is there.
Can it solve any cube, no matter
how much I mix it?
Theoretically it should. There are
43,252,003,274,489,856,000 different combinations. That’s about 43
quintillion. I have not tried all of them, but so far I have not found
one it couldn’t do.
Does it ever not solve a cube?
The colour system is a little finicky
with lighting so sometime RuBot has a little trouble reading the
colours. Sometimes RuBot's hands get out of alignment and he drops or
breaks a cube. This is especially true if the cube is very loose.
Is that all it does?
Is that not enough? There’s no
pleasing some people. Most of today’s robots are designed for a single
task. Even so, RuBot II could probably play chess very well.
Couldn’t you have built something useful?
No, I don’t want to build a dishwashing robot or a
laundry folding robot right now. I like to build robots that make
people smile. I like to build entertaining robots. Actually I think
that is useful.
If you change xxx it will be a
lot better…
Thanks for the suggestion but RuBot II is
finished. I might tinker a little bit but I have moved on to the next
project and will not be changing RuBot II any time soon.
Can I buy it?
RuBot is not for sale. It is the only one in the
world, a Guinness World Record holder and I get to travel the world with
it. That being said, everything has a price…
Can you build me one?
I could but I really don’t have the
time. If I was building another one, I would change a lot of things.
RuBot II was not built with the intention of anybody else operating it.
This means that it requires a lot of attention. A new model would
have to be more reliable and user friendly and this would require
development time. In short, yes but it would cost you a lot of
money.
Can I have the plans?
Sorry, I don’t have any plans. I have
to admit that it was all built on the fly.
Can you bring it to my
school/event/exhibition/TV show….?
Yes, bookings can be made through
the
mechatrons.com website.
How did you learn how to build robots?
I have been building things since I was a child.
I was lucky enough to get an apprenticeship when I left school and got
trained as a mechanic. Then I transferred to electronics and (after
many years) became an aircraft avionics inspector. I moved into flight
simulators and got some expertise in control systems. I also studied
part time and got degrees in Engineering and Computer Science.
What do you do for a living?
I am currently researching robotic
vision systems at Trinity College Dublin and working towards a PhD. I
also do robot display shows and special effects for movies and TV.
What else have you built?
This is a long list so I will restrict it to some
of the more interesting things. Most famously, I built two robots for a
BBC TV show called Robot Wars. These were called Dioitoir and Nemesis.
With them I won about 5 trophies over the years on the show. For
another BBC show called Techno games, I built an internal combustion
powered sprinting robot that set a new world land speed record for
petrol powered walking machines. I had a lot of fun building and
driving an all-terrain stock car for a Channel 4 TV show called Full
Metal Challenge. It was half Jeep/half dumper truck with articulated
steering. Other things include robotic bicycle, micromouse, trebuchet,
orange cannon and really the list goes on and on.
What are you working on next? I
am currently working on two projects. One is a robot vision system for
large aircraft and other vehicles. The other is a robotic TV
presenter. Watch this space…
|