Archive for April, 2010
From today and through this Blog I would like to start sharing with all of you the project I started 5 years ago and which is about to be released. In this Blog not only will I tell my adventures and hopes during this investigation but also every obstacle I found on the way. I will also share the robot evolution, from how the design idea arose to the problems I faced while making the casts. I would not like to forget all the people who for one reason or another have supported this project and specially those who are still in it and who have believed in it from the very beginning.
“WELCOME and remember that you should never stop being children”. Francisco Paz
CURRENT ROBOTICS AND THE NEED FOR AN OPEN-SOURCE ROBOT
I have been following for years the progress made in both the robotics and the artificial intelligence fields. In all this time (20 years) there are few projects which are still working, mainly due to the high costs involved in the making and investigation in this interesting but complex field. Of course, apart from robots such as ASIMO (http://world.honda.com/ASIMO/) or others with similar characteristics such as Toyota’s (http://www.toyota.co.jp/en/special/robot/), in which millions of dollars are invested every year, there are NO “real” alternatives for the ordinary consumer which allow to go into the robotics and the artificial intelligence world as a hobby.We could talk about small Japanese robots which are totally made up from servos and occasionally with small processors which let us do some (reduced) tasks such as walking, processing some images and/or talking.
At this point, where are we really? Nowadays we could perfectly make a difference between groups dedicated to creating robots (mechatronics) and those dedicated to doing research in the field of artificial intelligence. We could also divide the first one into two groups: big companies which make robots, especially mechatronically, with two clear targets: advertising and patenting their investigations, and on the other hand, small companies which make small devices (normally with servos) and sell them through the Internet at relative affordable prices for ordinary consumers.
Regarding the second group, those entirely dedicated to developing artificial intelligence, we can highlight Novamente LLC(http://novamente.net/), AGI Research Institute ( http://www.agiri.org/wiki/Main_Page), both managed by Ben Goertzel ( http://www.agiri.org/wiki/Main_Page) and Redwood Neuroscience Institute founded by Jeff Hawkins ( http://en.wikipedia.org/wiki/Jeff_Hawkins ) in 2002. Within this group there has been very important progress in Open Code dedicated to development and research of certain areas of artificial intelligence, paying special attention to the OpenCV project, developed by Intel about vision (colour and object detection, tracking, etc.) ( http://sourceforge.net/projects/opencvlibrary/) and in the field of speech synthesis and recognition (SRS- Speech recognition System), projects like Festival speech synthesizer (http://www.cstr.ed.ac.uk/projects/festival/) or the Sphinx speech recognition system (http://cmusphinx.sourceforge.net/html/cmusphinx.php) developed by Carnegie Mellon University.
According to Marvin Minsky’s words (co-founder of the MIT Artificial Intelligence Lab) pronounced at Boston University, “Artificial intelligence is brain-dead”. Minsky accused developers and rechearchers of making more mechatronic machines instead of creating “intelligent” machines able to carry out autonomous tasks. It is true that in the last years much advance has been made but these machines are not suitable for the market yet. Ben Goertzel also agrees with the idea that much of the research carried out regarding artificial intelligence is focused on solving specific problems such as more efficient research algorithms in search engines and not so much on developing a general purpose system.
Tomomasa Sato is one of the most prestigious engineers in the robotics field, holding the post of director of the Japanese Robotics Association. He has recently claimed that it is necessary to develope an Open Source Model-T robot in which all global standards may be applied to achieve a result as revulsive as Ford’s Model-T was for the car industry.
Could it be possible to do what Tomomasa Sato says? The answer is YES. However, there are many obstacles which make it difficult. The first one is economic since creating for example a structure in a SolidWorks programme takes too much time and requires deep knowledge about it. Patents from big companies limit small ones from embarking on that adventure because lawsuits, even if they are Open Source, could finish the project even before starting. Another obstacle would be giving a suitable design to the robot in order to create an important community around it and developing software for the robot. Creating a design implies making the casts and their costs would rise depending on how attractive we want it to be. The average rate for only one cast goes from 6.000 to 80.000 euros. If a small robot (60 cm) with an attractive design may need approximately 20 casts, the outlay to create the external structure could perfectly go from 400.000 to 600.000 euros excluding the internal parts of the robot (sensors, cams, PC board, memories, controllers, motors, servos, etc.)
iCub (http://www.robotcub.org/) is a 5 year old Open Source project which tries to make a real open code robot at a hardware and software level. You can watch some interesting videos about the project (http://www.youtube.com/watch?v=Dmj5TP7XIFE) or a video where iCub is trying to learn (http://www.youtube.com/watch?v=5jFfgJDbwxQ). Despite the project being entirely Open Source and being posible to access each one of the files which store the parts to make one ourselves, we are still far away from the fact that an ordinary user could do it, due to the lack of knowledge about how design programmes work and, above all, the high cost of all the pieces integrating the robot.
The design of a low cost open-source robot which integrates components currently found in the market at a low price (PC board, memories, ultrasound sensors, webcams, compasses, batteries, etc.), a plate frame with a suitable external design and a Linux operative system to integrate and handle every component through controllers “would be a revulsion similar to Ford’s Model-T in the past and Linux in the field of operative systems“.