Integration of soft-robotics and deep learning to assess the coordinated emission and reception dynamics in hipposiderid bats

Abstract

Some echolocating bat species with sophisticated biosonar systems such as the Old World leaf-nosed bats (Hipposideridae) change the shapes of their ultrasound emission (noseleaves) and reception baffles (pinnae) during the diffraction of the incoming and outgoing acoustic waves. Our prior research has shown that these noseleaf and pinna deformations are coordinated, but a functional significance for this coordination has yet to be demonstrated. Here, we combine a soft-robotic reproduction of the dynamic biosonar periphery in hipposiderid bats with deep neural networks to assess the impact of the coordination between the emission and reception dynamics. The biomimetic noseleaves and pinna are equipped with soft pneumatic actuators to accomplish life-like deformations. Control over this system allows us to reproduce not only the coordination patterns seen in bats but also other possible patterns that appear to be absent in bats. Comparing the suitability of these different coordination schemes between emission and reception could be used to assess which coordination pattern is best suited for a given task. As a first task, a direction finding paradigm has been selected for this assessment and work on identifying a network architecture that can take advantage of the dynamic nature of the echoes is currently in progress.