Engineers from the University of Oslo have developed a new robot which can teach itself to walk as it evolves through trial and error. The robot is called DyRET and is described in the study, “Self-Modifying Morphology Experiments with DyRET: Dynamic Robot for Embodied Testing.”
“Robots need to be able to adapt to complex and dynamic environments for widespread adoption, and adapting the body might yield more flexible and robust robots,” states the paper. “Previous work on dynamic robot morphology has focused on simulation, combining simple modules, or switching between locomotion modes. This paper presents an alternative approach: automatic self-reconfiguration of morphology on a four-legged hardware robot.”
A video was also provided with the paper to show the robot’s skills and to describe how it functions and adapts to new terrain. DyRET accomplishes this by shifting its body weight and taking different steps. It often falls over but this is also a step in the learning curve.
A range of motion sensors record the movements of the robot and allow DyRET to pick the most successful ones. It is designed in such a way that it does not move excessively when it is low on battery. The legs of the robot shorten at this time and it moves with shorter limbs to preserve energy.
This is a part of a new field called evolutionary robotics. It uses evolutionary computation to generate evolving robots capable of adapting to their environments through a process similar to natural evolution. If we look at the potential application of this field, it was far-reaching implications.
The reason why there has not been much progress in the field is because of the absence of standard research practices in the field. Scientists suggest potential avenues of research that could support “the establishment of evolutionary robotics as a canonical approach for the engineering of autonomous robots.”
Tonnes Nygaard is leading DyRET and he is a roboticist with Engineering Predictability With Embodied Cognition project at the University of Oslo. The project states its goal as “to exploit the form of various systems to develop predictive reasoning models as alternatives to traditional reactive systems.”