Smarter sensors, easier processing

Perceptual research in artificial intelligence usually focuses on the central processing of sensor data. The sensors (or sensor models) are often simplified and abstracted analogues of the sensorial systems of real animals. This approach places the burden of perception on the central nervous system. Moreover, it neglects the fact that sensors are important parts of the perceptual systems of animals integrated both with central processing and expressed behavior.

Indeed, as proposed by Rudiger Wehner (1987) and others, many sensorial systems found in nature have evolved to offload part of the perceptual processing to the periphery. Furthermore, some animal sensors have associated sensory behaviors that constrain or generate rich interactions that make perceptual interpretation easier, for example rat whiskers.

Specifically, through their physical interaction with the environment sensors select and preprocess data thereby substantially reducing the need for elaborate central processing. For example, insect photoreceptors provide the animal with a coarse image of the outside world in which details that are unimportant to the behavior are reduced and relevant features are enhanced.

In this workshop we want to bring together people working on smart perceptual systems that reduce the need for central processing both in animals and animats.