Evolutionary Morphology for Polycube Robots

Recent developments in robot technology have led to modular robots that are assembled from multiple parts. Modular robots have pros and cons. Their shape can be changed to suit a particular purpose and the same parts can be reused, which makes them fault-tolerant and extendable. However, such freedom and flexibility lead to higher cost so as to decide the configuration and actions to meet the purpose. With conventional robots, modular or non-modular, much research has been conducted on the way that a design is determined first and connections with sensory input, internal memory and motor output are acquired through learning. The designs are often based on existing creatures such as insects, dogs and earlier robot models, or parts of such creatures, for example, robot arms. As robot bodies operate in the real world, it makes sense to model them after living creatures that have been successful in nature. In fact, computer simulation patterns akin to living creatures have been created mainly in the field of artificial life, represented by the well-known L-System imitating plants and Dawkins’ biomorph (Dawkins, 1986) imitating insects. Sims’ virtual creatures (Sims, 1994a; Sims 1994b) feature not only a pattern but also a body and nerves, which are actually stylized art rather than robotics. Modular robots, which are easy to configure and recombine, have an advantage in that optimal robot shapes can be achieved, without modeling the existing creatures. Many existing modular robots are based on the block shape. Although automatic design research is not limited to the block shape, the shape has certain advantages: The basic units are easy to make, handle and combine. Furthermore, blocks are similar to cells, which enabled the simulation of biological development in the research on building structures. GOLEM Project (Lipson, 2000; Pollack, 2000) is one of the implementation models of Karl Sims experiments. They evolved frame-type virtual creatures using a linear actuator and fabricated them as a robot in a real world. In the modular robotics, new types of robots have been proposed with the functions such as a self-reconfiguration (Murata et al., 2002; Kurokawa et al., 2003) or self-reproduction (Zykov et al., 2005; Griffith et al, 2005). The LEGO Mind Storm is a famous block-based modular robot, parts of which are so small that half-assembled modules are combined in different placements (Lund, 2001). Ideally, robots can actively change their configuration by automatically recombining modules to fit a particular environment. However, this chapter deals with cube-shaped modular robots that can be manually recombined to achieve significant configurations that can adjust to actual environments using evolutionary morphology. This chapter describes how successfully EC