Sim2Real Transfer for Audio-Visual Navigation with Frequency-Adaptive Acoustic Field Prediction

arXiv:2405.02821v1 Announce Type: cross
Abstract: Sim2real transfer has received increasing attention lately due to the success of learning robotic tasks in simulation end-to-end. While there has been a lot of progress in transferring vision-based navigation policies, the existing sim2real strategy for audio-visual navigation performs data augmentation empirically without measuring the acoustic gap. The sound differs from light in that it spans across much wider frequencies and thus requires a different solution for sim2real. We propose the first treatment of sim2real …

abstract arxiv attention audio augmentation cs.ai cs.lg cs.ro cs.sd data eess.as measuring navigation policies prediction progress robotic simulation strategy success tasks transfer type vision visual visual navigation while