Share

How do robots see the world?

Future robots will evolve their behaviour depending on environmental conditions: this is one of Artificial Intelligence challenges. At ISTC the Laboratory of Autonomous Robotics and Artificial Life (LARAL) is contributing to the creation of a new generation of robots able to take initiatives and changing reality. 

Imagine you want to design a smart robot sensing the environment and carrying out difficult tasks. Traditionally, engineers tried to do this from the bottom up, building all the robot's parts and specifying their interactions rules in order to reach a desired behaviour. The result was always limited, since it was not possible to foresee everything. Now Artificial Intelligence is mostly giving up such hope of engineering control, relying instead on the so-called "self-organization". That is, letting the robot see the world with its eyes and learn from errors.

At ISTC the Laboratory of Autonomous Robotics and Artificial Life (LARAL) is walking this path. The idea works like this: the "designer" only sets up a flexible framework specifying the basic mechanisms of the robot – how it takes input from the environment and makes decisions. The following steps are left to evolution: the robot literary learns how to behave. After several experiences, it discovers the most useful properties and retains them in new tasks.

LARAL team uses the self-organizing method to develop coordinated groups of robots, which autonomously evolve. This approach is leading to a new generation of embodied agents that are able to directly interact with the physical world and with humans. Using a "personal" and not predetermined view of the world. 

Contact: Stefano Nolfi

ISTC Group: Laboratory of Autonomous Robotics and Artificial Life