Scaffolded Learning of Bipedal Walkers: Bootstrapping Ontogenetic Development

Abstract

Bipedal locomotion has several key challenges, such as balancing, foot placement, and gait optimization. We reach optimality from a very early age by using natural supports, such as our parents hands, chairs, and training wheels, and bootstrap a new knowledge from the recently acquired one. In this paper, we propose a scaffolded learning method from an evolutionary robotics perspective, where a biped creature achieves stable and independent bipedal walking while exploiting the natural scaffold of its changing morphology to create a third limb. Hence, we compare three conditions of scaffolded learning to reach bipedalism, and we prove that a performance-based scaffold is the most conducive to accelerate the learning of ontogenetic bipedal walking. Beyond a pedagogical experiment, this work presents a powerful tool to accelerate learning on robots.