Animat are artificial animals; the term is a contraction of "animal" and "materials" (and, coincidentally, also the third-person indicative present of the Latin verb animō which means to "animate, give or bring life"). The term includes physical robots and virtual simulations. The animat model includes features of a simple animal capable of interacting with its environment. It is, therefore, designed to simulate the ability to associate certain signals from the environment within a learning phase that indicate a potential for cognitive structure.
Animat research, a subset of Artificial Life studies, has become rather popular since Rodney Brooks' seminal paper "Intelligence without representation".
Development
The term was coined by S.W. Wilson in 1985, in "Knowledge growth in an artificial animal", published in the first Proceedings of an International Conference on Genetic Algorithms and Their Applications. Wilson's conceptualization built on the works of W.G. Walter, particularly his invention of the nuilt 2 three-wheeled sensor, propulsion motor for front-wheel drive vehicles. In Machina speculatrix, Walter introduced what can be described as a sub-animat, which chose actions based on needs and the sensory situation. A few rules were already introduced in this seminal work. There is, for instance, the linking of speeds of the two motors to the level of illumination. Norbert Wiener's theories postulated in the 1948 Cybernetics is also said to have inspired the simulation of animals, particularly the brain and behaviors of frogs (Rana computatrix), rats, and monkeys.
In its early conceptualization, the animats - was built as simple creatures and simulated behaviors, which pertain to genetic reproduction and natural selection. Wilson's animat, however, did not only interact with the environment but also learned from its "experience".
Theories and applications
An example using the Animat model as proposed by Wilson is discussed at some length in chapter 9 of Stan Franklin's book, Artificial Minds. In this implementation, the animat is capable of independent learning about its environment through application and evolution of pattern-matching rules called "taxons".
In 2001, Thomas DeMarse performed studies on 'neurally controlled animat'. Another recent development was the successful demonstration by Holland and Reitman of a rule-adaptive animat that has an optimized rate of satisfaction of two distinct needs.