A topological extension of movement primitives for curvature modulation and sampling of robot motion

Abstract This paper proposes to enrich robot motion data with trajectory curvature information. To do so, we use an approximate implementation of a topological feature named writhe , which measures the curling closed curve around itself, and its analog for two curves, namely linking number . Despite these features have been established their definition allows discrete calculation that is well-defined non-closed curves can thus provide information about how much line in space. Such lines be predefined by user, observed vision or, our case, inferred as virtual space curling. We augment encapsulated Movement Primitive (MP). propose method determine many segments best characterize then find such segments. results generative model permits modulating generate new samples, while still staying within dataset distribution being able adapt contextual variables.