Experiences from Real-World Evolution with DyRET: Dynamic Robot for Embodied Testing

Abstract

Creating robust robot platforms that function in the real world is a difficult task. Adding the requirement that the platform should be capable of learning, from nothing, ways to generate its own movement makes the task even harder. Evolutionary Robotics is a promising field that combines the creativity of evolutionary optimization with the real-world focus of robotics to bring about unexpected control mechanisms in addition to whole new robot designs. Constructing a platform that is capable of these feats is difficult, and it is important to share experiences and lessons learned so that designers of future robot platforms can benefit. In this paper, we introduce our robotics platform and detail our experiences with real-world evolution. We present thoughts on initial design considerations and key insights we have learned from extensive experimentation. We hope to inspire new platform development and hopefully reduce the threshold of doing real-world legged robot evolution.

Publication
In 2019 Symposium of the Norwegian AI Society (NAIS19), Springer CCIS.