That Sounds Right: Auditory Self-Supervision for Dynamic Robot Manipulation
This addresses the problem of contact-rich, dynamic manipulation for robotics by leveraging sound to overcome limitations in visual and tactile sensing, though it is incremental as it builds on existing self-supervised methods.
The paper tackles the challenge of learning dynamic robot manipulation by using sound as a sensory input, collecting 25k interaction-sound pairs and applying self-supervised learning to improve behavior prediction, resulting in a 34.5% lower MSE than supervised learning and an 11.5% improvement in audio similarity metrics on a robot.
Learning to produce contact-rich, dynamic behaviors from raw sensory data has been a longstanding challenge in robotics. Prominent approaches primarily focus on using visual or tactile sensing, where unfortunately one fails to capture high-frequency interaction, while the other can be too delicate for large-scale data collection. In this work, we propose a data-centric approach to dynamic manipulation that uses an often ignored source of information: sound. We first collect a dataset of 25k interaction-sound pairs across five dynamic tasks using commodity contact microphones. Then, given this data, we leverage self-supervised learning to accelerate behavior prediction from sound. Our experiments indicate that this self-supervised 'pretraining' is crucial to achieving high performance, with a 34.5% lower MSE than plain supervised learning and a 54.3% lower MSE over visual training. Importantly, we find that when asked to generate desired sound profiles, online rollouts of our models on a UR10 robot can produce dynamic behavior that achieves an average of 11.5% improvement over supervised learning on audio similarity metrics.