Framsticks

Research in complex collective and multi-agent systems often involves building models of three-dimensional biological life or evolving such structures in virtual environments. Applications stemming from evolutionary design, engineering, robotics, and artificial life require processing of large numbers of such agents that are encoded in some form of a "genotype". However, what is important in evaluation is the "phenotype", i.e. the actual 3D body and its properties. This work introduces a number of ways in which a measure of similarity of 3D agents can support researchers in recognizing the link between the genotype and phenotype spaces, building taxonomies of 3D bodies and automatically selecting representative agents. The measure of similarity employed here is based on phenotypes and places few restrictions on the compared designs, so it can be applied independently of genetic representation.