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 'Audio Robot Learning' (AuRL) 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.
We collect a dataset of 25k interaction-sound pairs across five dynamic tasks using commodity contact microphones placed on and around the robot.
To learn behaviors with AuRL, we first transform the raw audio waveform into a Mel spectrogram. Then, to learn good representations, we use self-supervised learning on our audio data using the BYOL algorithm. Finally, we train a linear model to predict the behavior primitives on top of the self-supervised representations by minimizing the MSE loss between the predicted actions and the test action using simple supervised training.
Our key results are as follows :