Scientists have tested a truly revolutionary theory that sees the robots “think” the actions to be performed with a given object. One could therefore assume that the robots are able to learn from their own observations and experiences. This latest discovery is the result of the project PACO-PLUS (Perception, action and cognition through learning of object-action complexes), funded with 6.9 million € in reference to the theme “Information Society Technologies” the Sixth Framework Programme (FP6) of the EU.
The project partners have sought to test the theory known as “object-action complexes (OACs). The OAC units are characterized by “thinking by doing”, an approach used for the design of software and hardware that allows the robots to think of the objects compared to actions that can be done with them. For example, if a robot sees an object with a handle could grab it. However, if the object has an opening, the robot could enter a different object or pour the liquid. If, again, the object is fitted with a lid or a door, the robot could open it. Thus the objects become important compared to the number of shares that a robot could perform.
This theory opens up a lot more interesting scenario regarding the possibility that robots think independently, because – according to the project partners – increases the odds that the robots take on what is called “emergent behavior” or that spontaneously develop complex behaviors on the basis of rather simple rules.
The approach chosen by the team plays in many respects the learning processes of children. These, in fact, when you are faced with a new object, try to grab it, eat it or beat it against another object: it is just experimenting and error (eg trying to insert an object into a round hole square) that are able to gradually expand the number of its shares. A small child also learns from the people next to him.
Within the project PACO-PLUS most of the experiments were carried out on humanoid robots, or robots with human features. “The humanoids are personifications artificial perceptual and motor skills varied and complex, making them [?] The best platform to study the cognitive processes and cognitive processes of information processing,” explained Dr. Tamim Asfour group research on humanoid Department anthropometry Institute of Technology in Karlsruhe (Germany), and the project coordinator.
Dr. Asfour said the team did their job in the wake of what has been done by Rodney Brooks, one of the leading professors of robotics, which now collaborates with the Massachusetts Institute of Technology (MIT), USA. As he explained, Professor Brooks was the first to state clearly that cognition is a function of our perceptions and our ability to interact with the surrounding environment. In other words, cognition is the result of our physical presence situational environment. ”
“Professor Brooks believed that gaining the ability to move in the environment, as well as interact, represent a problematic aspect of biological evolution. However, he believed that if a species acquired these skills, the process of development of symbolic reasoning abstract thinking was pretty simple, “explained Dr. Asfour. The opposite approach is that of artificial intelligence to the effect that if we can create a form of intelligence developed enough, the machines should be able to perceive and solve problems.
It is unclear which of the two theories is correct, and the researchers themselves admit that although progress has been made there is still no robot with “naturally” of intelligence. “Of course the interpretation of Hollywood is still far, but applications and demonstrations of the project PACO-PLUS show that perhaps we are on the right path,” said team.
