Verifiable Learned Behaviors via Motion Primitive Composition: Applications to Scooping of Granular Media

TL;DR

This paper presents a framework for generating verifiable robotic behaviors from natural language inputs by composing motion primitives, demonstrated through simulation and real-world granular media scooping tasks.

Contribution

It introduces a natural-language behavior abstractor that synthesizes behaviors from motion primitives, ensuring probabilistic verifiability of the generated behaviors.

Findings

Behaviors can be reliably generated from natural language inputs.

The framework ensures behaviors are probabilistically verifiable.

Successful demonstration on both simulation and hardware.

Abstract

A robotic behavior model that can reliably generate behaviors from natural language inputs in real time would substantially expedite the adoption of industrial robots due to enhanced system flexibility. To facilitate these efforts, we construct a framework in which learned behaviors, created by a natural language abstractor, are verifiable by construction. Leveraging recent advancements in motion primitives and probabilistic verification, we construct a natural-language behavior abstractor that generates behaviors by synthesizing a directed graph over the provided motion primitives. If these component motion primitives are constructed according to the criteria we specify, the resulting behaviors are probabilistically verifiable. We demonstrate this verifiable behavior generation capacity in both simulation on an exploration task and on hardware with a robot scooping granular media.