The Age of Thinking, Self-Developing Robots Has Finally Arrived

Institute of Robotics in Scandinavia (iRobis) has announced commercial availability of Brainstorm®, the world’s first “complete cognitive software system for robots”. The system turns robots into self-developing, adaptive, problem-solving, “thinking” machines. The system automatically writes control programs for any robot on which it is installed, dramatically shortening development time and cost. The same technology is used to allow robots to adapt to new circumstances and solve other problems while in operation.

The coming of such creatures has been predicted for decades. But how long does it really take to put thinking, self-developing robots on the market? This question was addressed recently at the RoboDevelopment Conference in Santa Clara, California. It really depends on when you start.

The first scientifically recorded effort to get a modern computer to learn took place in 1958. So if you started then, and guessed it would take about half a century, you would have been right. It didn’t happen all of a sudden. The first mobile robot to be controlled by artificial intelligence was created in 1966. By the 1980s, machine learning was widely recognized as a distinct scientific discipline with many disciples.

The secret ingredient in iRobis’ software is a method called “genetic programming” (GP). A number of scientists since the time Charles Darwin published On the Origin of Species to the present day have suggested that human thought and problem solving follow the same pattern as natural evolution. Connections between simpler thoughts that make less sense together do not survive the process. Stronger relationships do, as part of the evolution of more complex thoughts and solutions. GP uses this idea to evolve arbitrarily complex computer programs.

GP developed out of the rich and varied research on machine learning during the 1980s and found its special defining moment in a 1992 book, Genetic Programming: On the Programming of Computers by Means of Natural Selection by Stanford University researcher John Koza. According to the textbook, Genetic Programming An Introduction, by 1998 more than 800 papers had been published related to GP by 200 authors.

One of those authors was Peter Nordin, the human brain behind Brainstorm. Dr. Nordin began publishing AI articles in 1992 and followed with an impressive run of systematic research on GP and robotics. In 1999, he presented An Evolutionary Architecture For A Humanoid Robot, a new GP based software architecture for robots. From 1998-2003 he ran The Humanoid Project at Chalmers University of Technology in Sweden. Hundreds of robots of various types and sizes were created and the GP approach got plenty of exercise. That led to a contract with the Swedish Defence Department in 2006 to develop a commercial version of the software and the creation of iRobis, his 19th company to apply GP to a commercial product.

So if you’re Peter Nordin, starting work in the early 1990s, dedicating your career to GP and intelligent robotics, you can say that it takes 15-20 years to produce commercial software for thinking, self-developing robots – as well as sufficient funding of course.

His 1999 paper presented results of experiments carried out on a humanoid robot known as Elvis. The successful experiments included balancing, walking, bifocal vision, navigation, audio orientation, and object manipulation. Elvis’ hips moved appropriately when he danced, which may be the reason for his name. Since that time there have been improvements in the speed of evolutionary processing, the use of the robots’ “imagination” for planning, adapting, and problem solving, and advanced physical control such as hand-eye coordination among other things.

How long will it take to bring new “thinking,” self-developing robots to market? Brainstorm is the first tool for development of GP robotics system on the market. The development of a tool suggests a high level of technical maturity. How long will it take product developers to create robots using the tool?

An answer is suggested by the work of Almir Heralic, a former student of Peter Nordin turned professor. He created Humanoid Robot HR-2 using software implementing the GP architecture during a three-month project at Chalmers. At a cost of only $300, it rates quite well in the advanced robotics field. And the video documentation at the HR-2 website is pretty good too.

The announcement does not mean that Brainstorm is available to download for a free 30-day trial. In the short-term, iRobis expects to work directly with early commercial adopters and researchers to create prototypes with previously unseen levels of intelligent autonomous behavior and to prove the value of the system for rapid development and advanced experimentation. The search is on for high-value strategic partnerships.

Contact: Brainstorm Development, Sales, Marketing, and Partnerships