A new kind of cognitive autonomous robots
European computer scientists have developed a new kind of cognitive robots. In two articles, ICT Results describes how the researchers have combined classical rule-based artificial intelligence and artificial neural networks (ANNs) (links to part 1 and to part 2). This learning robot 'employs ANNs to manage the low-level functions based on the visual input it receives and classical AI to serve as a supervisory mechanism.' So far, this robot can solve puzzles by itself and other complex tasks with no additional programming. The project has been so successful that the European Union decided to fund 13 additional projects based on these results. But read more...
You can see on the left how such a system can learn by itself. In this example, it is used to solve puzzle problems. Here is what you can see from top to bottom: the system starts with selecting one object (the circle) and moves it to the corresponding hole; the system then selects the arch, which cannot be put into the corresponding hole, since it is occluded by the square; the system puts the arch back to its original position and continues with another object: after the triangle and the semi-circle, the system selects the square, such thatthe occlusion of the arch-hole is removed: finally, the arch is put in the correct hole and the puzzle is solved. (Credit: Linköping University, Sweden)
This figure has been extracted from a paper presented at the Association for the Advancement of Artificial Intelligence (AAAI)'s 2005 Fall Symposium held in Crystal City, Arlington, Virginia, in a session focused on From Reactive to Anticipatory Cognitive Embodied Systems.Here are two links to the abstract and to the full paper, A COSPAL Subsystem: Solving a Shape-Sorter Puzzle (PDF format, 5 pages, 1.13 MB).
The COSPAL project -- short for "COgnitive Systems using Perception-Action Learning" -- was funded by the European Union. It was coordinated by Michael Felsberg, who works at the Computer Vision Laboratory at Linköping University (LiU), Sweden. He worked closely with professor Goesta Granlund, head of the Division of Computer Vision at LiU.
The cost of this project, completed in 2007, was €2.35 million. It was apparently so successful that 13 other projects with LiU participation have been awarded funds from the EU Seventh Research Framework Programme (FP7). Here is the list of these programs.
Here is why the researchers decided to combine two AI approaches. "'Developing systems in classical AI is essentially a top-down approach, whereas in ANN it is a bottom-up approach,' explains Michael Felsberg. 'The problem is that, used individually, these systems have major shortcomings when it comes to developing advanced ACS architectures. ANN is too trivial to solve complex tasks, while classical AI cannot solve them if it has not been pre-programmed to do so.'"
And Felsberg's team found that using the two technologies together solves many of those issues. "'In this way, we found it was possible for the robots to explore the world around them through direct interaction, create ways to act in it and then control their actions in accordance. This combines the advantages of classical AI, which is superior when it comes to functions akin to human rationality, and the advantages of ANN, which is superior at performing tasks for which humans would use their subconscious, things like basic motor skills and low-level cognitive tasks,' notes Felsberg."
The second ICT Results article is more focused on how a "trial and error approach could lead to more autonomous robots and even improve our understanding of the human brain." For example, Granlund noted that children are "'always testing and trying everything' and by performing random actions -- poking this object or touching that one -- they come to understand cause and effect and can apply that knowledge in the future. By experimenting, they quickly find out, for example, that a ball rolls and that a hole cannot be grasped. Children also learn from observing adults and copying their actions, gaining greater understanding of the world around them."
Anyway, don't worry. These learning robots are not even challenging children -- at least in 2008. "Though a learning, cognitive robot of the kind developed in COSPAL constitutes an important leap forward toward the development of more autonomous robots, Felsberg says it will be some time before robots gain anything close to human cognition and intelligence, if they ever do. 'In human terms, our robot is probably like a two or three year old child, and it will take a long time for the technology to progress into the equivalent of adulthood. I don’t think we will see it in our lifetimes,' he says."
Sources: ICT Results, March 21 & 27, 2008; and various websites
You'll find related stories by following the links below.