Morphology, control and passive dynamics

Traditionally, research in robotics has focused primarily on control. However, an increasing body of research supports the notion that the structure and mechanical characteristics of the robot body, that is its morphology, play a crucial role in behavior generation and control. The morphology not only determines the kinematics and dynamics of the robot, and thereby the possible repertoire of behaviors, but also affects the control required for these behaviors. A well designed morphology can lead to drastic reductions in control requirements, as well as improved controllability. On the other hand, a poorly designed morphology may lead to low controllability, require complex control algorithms, or in the worse case, simply be inadequate for the task. The benefits of good morphological design have become apparent in robotics through the design of robots which utilize passive dynamics. Starting with the work of Tad McGeer who built self-stabilizing passive mechanisms which could walk down a slope in the absence of control, numerous robots have been developed by Steve Collins, Andy Ruina, Russ Tedrake, Richard van der Linde, Martijn Wisse and others demonstrating how well-designed morphologies can lead to reduction in control requirements and improved efficiency. The relevance of this idea extends beyond the design of passive dynamic walkers to the design of running robots (Marc Raibert, Hiroshi Kimura, Jorge Cham, Fumiya Iida, Tao Geng), climbing robots (Metin Sitti), underwater robots (Edward Colgate, Michael Epstein, Malcolm MacIver) and manipulators (Suguru Arimoto). The relationship between morphology and control requirements has been characterized by Rolf Pfeifer as the “morphology and control trade-off”, making explicit the notion that an intrinsic relationship exists between morphology and control, and that control can be traded off for morphological properties. The discovery that the morphology can perform computation, also known as morphological computation (Chandana Paul), illuminates the main mechanism behind this trade-off. It shows that computational processes involved in control can be directly subsumed by the morphology, further corroborating the strong link between morphological design and control complexity. In addition to the design of the morphology, the interaction between morphology and control also plays an important role in behavior generation. The origins of this idea can be traced back to the work of cyberneticists such as Ross Ashby and Grey Walter who demonstrated that behavior could emerge