Morphological Computation - Nonlinear Materials for Nonlinear Control, Sensing and Computation

With the rise of soft robotics the focus of the design process for machines has shifted to nonlinear materials and clever mechanisms. The underlying idea is to exploit the often nonlinear dynamics of compliant morphologies to implement beneficial functionalities. This includes the physical embodiment of control, sensing, computation and even learning. This is partly motivated by the large number of examples from Nature, where we can see that biological systems rely heavily on their body morphology to survive in and interact with a complex, noisy and often unpredictable world. Morphological features endow biological systems with impressive levels of energy efficiency, robustness, resilience, and the ability to learn in complex environments. The approach to transfer these capabilities to intelligent machines through the use of clever morphologies is often referred to as morphological computation. Opposed to classical robotics, which actively suppresses any nonlinear and complex dynamics in bodies to facilitate modeling and control, morphological computation proposes to actively embrace and exploit nonlinear dynamical features. This implies a radical shift of how we should design and built machines. In order to be able to realize the full potential of morphological computation we need better materials, more intelligent structures, and novel mechanisms that allow us to design dynamic and responsive morphologies.

With the help of examples we will show how nonlinear dynamical properties can be exploited to implement useful functionalities for robots, including for control, sensing and computation. I will further discuss the requirements for the next generation of materials needed for morphological computation and the future of the field going from static bodies to adaptive morphologies and, eventually, to growing machines.