Department of Computer Science Department of Computer Science
An agent usually is seen as an essentially distinguished subentity of its environment, being able to act in it in an autonomous fashion. The environment can exert its influence on the agent in a manifold variety of different ways. Sensors and actors constitute the interface between environment and agent, enabling the latter to identify its current state and, using this knowledge, to gain certain control over its own fate. Since information is central for the ability to perform effective actions, an efficient utilization of available state information is vital to survive in an environment governed by external pressure selection. Managing this ecology of information is the central role of the perception-action loop of biological agents.
The principle of accessing new sensory channels and to make use of the implicit information coded there has been exploited in nature to a formidable degree. Olfactory, tactile, auditive and visual, but also e.g. electric and even magnetic senses have been realized in a vast multitude of variants, often utilizing organs not originally ``intended'' for the purpose they serve. At the same time, as already Darwin points out, it seems difficult to believe that an organ of such complexity as e.g. the eye ``[...] could have been formed by natural selection [...]''. Nevertheless photoreceptors of widely different degree of differentiation have evolved in at least forty independent lines of descent.
The study of sensor evolution is interested in understanding the mechanisms that drive the creation of sensors exploiting information sources in such a large variety of ways. We will relate the it to recent theories about the emergence of variance in the fossil record due to the emergence of new perceptual niches, as well as to understanding how living beings make optimal use of the available sensomotoric equipment. A second is understanding how agents can adapt themselves with a higher degree of not just sensoric, but neural flexibility to their environment.
Some issues lie even more at the core of Artificial Life research: recent research found intimate connections with questions of evolvability. In addition, there are strong indications that understanding the acquisition and processing of sensoric information by an agent can provide deep and far-reaching insights into:
By the end of the tutorial:
The targeted audience includes a broad range of people, in particular, people interested in: general evolutionary mechanisms for sensor development, also with respect to perspectives for hardware evolution; general questions of (embodiment) evolution and evolvability; fundamental questions of information processing in embodied perception-action loops and its role for modelling and understanding life-like systems.