Many robots solve complex tasks in closed environments that are fully known by the robot designer. Robots are much rarer in our every day environment. The main reasons are its complexity and openness that frequently result in unpredictable situations created by new objects or new dynamics of interaction. Numerous applications of robotics would benefit from robot abilities to deal with open environments. A lot of research is focused on this topic, but a question remains unanswered: how to evaluate the ability of a robot to face new contexts? The performance of a robot in a particular environment says little about what this performance will be in a different environment, even if it seems similar. A robot programmed to manipulate boxes may not be able to manipulate balls of the same size. Likewise, a robot perception system may be fooled by a change of luminosity. Consequently, the performance to expect in a new situation cannot be systematically deduced from what has been observed in known situations. There is then a need for a criterion that allows to compare approaches with respect to their ability to endow robots with the robustness or adaptive abilities to deal with new situations. To this end, it is proposed to define criteria based on the notion of creativity. A definition is given in the context of robotics and examples of use for learning and developmental processes are given.
Pour citer cet article :Doncieux Stéphane (2016/1). Creativity: a Driver for Research on Robotics in Open Environments. In Khamassi Mehdi & Doncieux Stéphane (Eds), New approaches in cognitive robotics, Intellectica, 65, (pp.205-221), DOI: 10.3406/intel.2016.1795.