A multi-level robotic architecture for biologically-inspired modeling

Traditionally, modeling of neurobiological systems has involved development of computer-based simulations. As opposed to physical experimentation, simulations tend to over-simplify environmental conditions. Yet, in many cases such environmental conditions are critical to experiment outcome. In the case of animal behavior, simulation-only arenas can serve as a preliminary platform for model experimentation. Realistic physical environments are required for final evaluation of model correctness. In this paper we present our work with physical robots as testbed for animal behavior experimentation under realistic environmental conditions.