Old problems are still there: robot intelligence as the interplay between many components of complex architectures

Abstract

The goal of the two Special Issues about “Artificial Intelligence techniques for Robotics: sensing, representation and action” is to present, discuss and assess recent advances in the use of Artificial Intelligence methods to build better robotic systems. When we say “better”, what are we talking about? Nowadays we know that elephants do not play chess [1] and, despite the astonishing performance of IBM Watson in Jeopardy [2], we argue that robot intelligence is a property of complex robot systems interacting with real world environments (including the robot itself). The question here is how can we achieve a level of robot performance that can, even loosely, resemble what any person can see in such movies as “Robot and Frank” [3]? As a matter of fact, as we expect robots to leave lab settings and to take part to our everyday life, either at home or in public spaces [4], the concept of AI as a separable and somewhat peculiar part of robot architectures shows its intrinsic limitations. The stress on sensing, representation and action of this Special Issue is motivated by the need of