Designing complex responses in smart structures

Coming up with increasingly smart robotic structures requires computationally solving elastic inverse problems; letting an algorithm find a geometry that results in the desired mechanical response. These approaches are limited by our ability to describe the desired structural performances, i.e. tell the inverse algorithm what we want our structure to do. For example, if we describe the desired behaviour using an elasticity tensor, we will be limited to linear-elastic structures — no matter our computational resources and the quality of our inverse algorithms. Overcoming this limitation requires coming up with novel ways to describe structural performances. In this talk, I will present two approaches towards designer structures with advanced responses. In the first approach, that we call design by code, the desired behavior is expressed as a source code in a 'structural programming language'. In the second approach, called design by data, the desired behaviour is described by a set of examples of desirable and undesirable responses. These approaches allow unprecedented structural performances ranging from the realisation of a fully mechanical 8-bits CPU, to metamaterials that selectively react to verbal commands.