A Brief(ish) History of Rodney the Robot
by creator Bob Mottram

What next?
So, where is all this headed?  Is there any point to building little humanoids, or would
I be better off spending more time in the pub?  I think what I've done so far has demonstrated to me just how complex
we humans are, and how much we take
for granted.  Many things which we can achieve without effort, such as recognising everyday objects and judging how far away they are, are actually extremely complex feats of sensing and bioengineering.  Despite decades of Moore's law and the prophecies of countless AI luminaries the sophistication of biological systems still far outstrips anything that we can attain with computer technology.  But this isn't any reason to be downheartened.  The other thing which development on the robot has shown is that there are solutions to some of the difficult problems involving visual perception, but that these are usually obtained through a great deal of programming and testing effort and they usually require the combination of multiple different algorithms working together in harmony.  In the past many AI researchers have tended to think that there was some kind of instant catch-all solution to the problem of building an intelligent entity – a notion which came to be known as “physics envy”.  But based upon what I've done so far with Rodney I don't think that there is any such solution in terms of a one line equation which could be applied carte blanche.

My aims for the future will be firstly to consolidate the systems which have been developed already.  This means making them more reliable and more modular, so that they really become useful building blocks for different sorts of robot.  In my opinion there is nothing worse than fragile “flaky” software which only works under extremely specialised or unrealistic conditions and all too often passes as AI research.  Breaking the huge monolithic mass of code, which used to be Rodney's software, into far more manageable chunks (DLL files) has helped a great deal to simplify the overall software design.  I may also create linux versions of the software, because linux is increasingly becoming a popular operating system for robotics applications.One critical omission from the robot at present is that all-important quality – learning.  Most of the software systems developed to date are fixed hand crafted solutions with little or no ability to change as the result of experience.  I havn't quite decided upon how I'm going to add learning, but my initial thoughts are a couple of extra modules which will allow the robot to learn the faces of people that it meets and to have a spatial awareness module which would keep track of the positions and types of object within the robot's immediate vicinity.

Many AI researchers are obsessed with making predictions, but I've worked in the software industry for over a decade so I know just how hard it is to predict what's going to come next.  I avoid making predictions, and I'm not entirely sure where the Rodney project will end up.  It's a bit of a journey into an unknown land.  Most of my motivation for doing a project like this is just to try to satisfy my own curiosity to see how far it's possible to take the technology.  Is it possible to build a machine which is similar to a person?  Can you build a robot that could be said to have a “personality”, a personal history and autobiographical memory?  I suspect that the answer is “yes”, but only time and some engineering effort will tell.  If I get to the stage where Rodney can do some reasonably interesting or fun things then I may consider marketing it as an interactive toy in the style of the Sony Aibo robot dog.  But for the present Rodney remains merely an interesting mechanical curiosity.

1 2 3