As mobile robotics technology continues to advance, the manual development of algorithms and controllers for these systems will become less feasible, if possible at all. As such, it will become increasingly necessary to turn to techniques that allow the automatic design of such systems, both in software and indeed in hardware. Evolutionary methods can provide powerful tools for automatic design as is evidenced by the abundance of diverse natural systems from the simple to the massively sophisticated and coupled. One of the fundamental features, and one of the lesser understood phenomena, in biology is that of speciation. In order to better understand the development and creation of species, and their role in evolution, a method for tracking speciation in simulation is presented. While there is much dispute in the field of biology as to the precise definition of the term species, there is little debate that the natural world is full of distinct subpopulations. Each of these populations has developed unique features for, and solutions to, the problem of surviving in an incredibly complex world. The power of investigating species over individual agents arises from improved robustness of a grouping of like individuals as opposed to single entities more sensitive to very local conditions and interactions. In essence, a species is a more complete view of the fitness and survivability of a genome than a single agent. A simulation engine is presented which allows the study of evolution from a species point of view.; Results from simulations offer insight into the role of genetic neutrality in evolution as well as the effect of this neutrality on mechanisms such as mutation pressure. Several results are presented which provide insight into these features which can serve as analogs for similar biological effects as well as features such as genetic cross drift (or convergent evolution of species). Some insight into the such as solution bloating are also detailed.; The simulation engine presented combines these and other features to serve as a tool for the investigation of the evolution of agents as an artificial process as well as a natural one.
展开▼
