Framsticks

Life is one of the most complex phenomena known in our world. Researchers construct various models of life that serve diverse purposes and are applied in a wide range of areas — from medicine to entertainment. A part of artificial life research focuses on designing three-dimensional (3D) models of life-forms, which are obviously appealing to observers because the world we live in is three dimensional. Thus, we can easily understand behaviors demonstrated by virtual individuals, study behavioral changes during simulated evolution, analyze dependencies between groups of creatures, and so forth. However, 3D models of life-forms are not only attractive because of their resemblance to the real-world organisms. Simulating 3D agents has practical implications: If the simulation is accurate enough, then real robots can be built based on the simulation, as in [22]. Agents can be designed, tested, and optimized in a virtual environment, and the best ones can be constructed as real robots with embedded control systems. This way artificial intelligence algorithms can be “embodied” in the 3D mechanical constructs.

Perhaps the first well-known simulation of 3D life was Karl Sims' 1994 virtual creatures [34]. Being visually attractive, it demonstrated a successful competitive coevolutionary process, complex control systems, and interesting (evolved) behaviors. However, this work did not become available for users as documented software. A number of 3D simulation engines was developed later (see [36] for a review), but most of them are either used for a specific application or experiment (and are not available as general tools for users) or focus primarily on simulation (without built-in support for genetic encodings, evolutionary processes or complex control). Notable exceptions are the breve simulation package described in Chapter 4 and a recent version of StarLogo — StarLogo TNG (Chapter 6).