ARTIFICIAL PIGS IN SPACE - USING ARTIFICIAL-INTELLIGENCE AND ARTIFICIAL LIFE TECHNIQUES TO DESIGN ANIMAL HOUSING

Computer simulations have been used by us since the early 1970s to gai n an understanding of the spacing and movement patterns of confined an imals. The work has progressed from the early stages, in which we used randomly positioned points, to current investigations of animats (com puter-simulated animals), which show low levels of learning via artifi cial neural networks. We have determined that 1) pens of equal floor a rea but of different shape result in different spatial and movement pa tterns for randomly positioned and moving animats; 2) when group size increases under constant density, freedom of movement approaches an as ymptote at approximately six animats; 3) matching the number of animat s with the number of corners results in optimal freedom of movement fo r small groups of animats; and 4) perimeter positioning occurs in grou ps of animats that maximize their distance to first- and second-neares t neighbors. Recently, we developed animats that move, compete for soc ial dominance, and are motivated to obtain resources (food, resting si tes, etc.). We are currently developing an animat that learns its beha vior fi om the spatial and movement data collected on live pigs. The a nimat model is then used to pretest pen designs, followed by new pig s patial data fed into the animat model, resulting in a new pen design t o be tested, and the steps are repeated. We believe that methodologies from artificial-life and artificial intelligence can contribute to th e understanding of basic animal behavior principles, as well as to the solving of problems in production agriculture in areas such as animal housing design.