Genetic robot morphology tool for interactive evolutionary design: using human judgement to improve appearance of social robots.

Project Details


This project proposes a novel method to improve the appearance of any physical product, focusing in particular on social robots. Social robot’s aesthetic has indeed a direct influence on its ability to create and maintain interactions with humans. But, as robot’s appearance can only be evaluated by humans, there are so far no direct methods to rigorously and efficiently improve it. Therefore, we propose to use human judgments as the fitness function of a genetic algorithm (GA, i.e. a Darwinian iterative process) coding for the robot appearance.

The first part of the project aims at creating a genetic robot morphology tool (named GEROMOT) that will link any parametric design to a genetic code. The novelty is that all solutions generated by the algorithm will take into account physical constraints, and thus be realizable.
In the second part, the purpose is to develop tools to rigorously collect human judgments as part of the GA process.
Finally, the last part aims at evaluating the efficiency of GEROMOT, in order to show that improving virtual appearance of social robots will result in improving their actual physical appearance. In particular, we will test the hypothesis that such improvement is correlated with better Human Robot Interaction (HRI).
Effective start/end date1/10/1730/09/21


  • robot morphology

Flemish discipline codes

  • Biomaterials engineering not elsewhere classified