| On 2019-05-21 at 13:30:00 (Brussels Time) |
Abstract
"Flora robotica" is an EU Horizon2020 FET project that investigates the hybridization of plants with self-organizing robots, for the application of construction. I will present the aspects of the project on which I have worked, mainly during my doctoral research. The talk will cover work in three main categories. The first category will be the control mechanisms for autonomous steering and shaping of natural plants. This is done by using robotic provision of stimuli to trigger natural phototropism (as a kind of bio-hybrid actuation), and by using computer vision techniques to track the plant in a closed-loop experiment setup. An evolutionary robotics approach to controlling plant growth and motion is used for the task of hitting both arbitrary and adaptive targets. A machine learning approach is then integrated, and used for the task of obstacle-avoidance. The second category will be the development of distributed bio-hybrid hardware for the autonomous control of plants. This is applied to growth of user-defined patterns in climbing plants on static mechanical scaffolds. The third category will consider the cases in which these mechanical scaffolds would be built by swarm construction or by adaptive self-assembly, using off-the-shelf mobile robots as experiment stand-ins. An ad-hoc approach to self-organized paths is used in multi-robot manufacturing of a particular structural system, one that displays high geometric nonlinearity. A bioinspired generative controller, developed within "flora robotica," is used for adaptive self-assembly in mobile robots. It is also used within a preliminary interactive evolution setup, for non-expert users to steer self-organized construction.
Keywords
bio-hybrid, swarm construction, evolutionary robotics, natural plants
