Whisker-Inspired Tactile Sensing: A Sim2Real Approach for Precise Underwater Contact Tracking

Abstract

Aquatic mammals, such as pinnipeds, utilize their whiskers to detect and discriminate objects and analyze water movements, inspiring the development of robotic whiskers for sensing contacts, surfaces, and water flows. We present the design and application of underwater whisker sensors based on Fiber Bragg Grating (FBG) technology. These passive whiskers are mounted along the robot$'$s exterior to sense its surroundings through light, non-intrusive contacts. For contact tracking, we employ a sim-to-real learning framework, which involves extensive data collection in simulation followed by a sim-to-real calibration process to transfer the model trained in simulation to the real world. Experiments with whiskers immersed in water indicate that our approach can track contact points with an accuracy of $<2$ mm, without requiring precise robot proprioception. We demonstrate that the approach also generalizes to unseen objects.

Figures (9)

• Figure 1: (A) Robot arm instrumented with a whisker sensor array in a crowded fish tank. (B) A close view of the whisker array in water. (C) Predicted contact positions when whiskers sweep over 3 objects.
• Figure 2: Overview of our sim2real learning framework. Moments computed in simulation are processed and used as inputs for training WhiskerNet, which is optimized using mean squared error (MSE) loss to predict contact locations. A calibration process, leveraging Gaussian Process Regression, maps wavelength data to base moments in simulation and real-world scenarios. The trained WhiskerNet model is then deployed in the real world for accurate contact location prediction.
• Figure 3: (a) Schematic of the whisker sensor design and components. (b) Image of the fabricated and encapsulated whisker sensor.
• Figure 4: Finite element analysis of the bridge structure in the whisker sensor design, illustrating the distribution of equivalent strain across the structure. The color gradient represents strain magnitude, with red denoting regions of higher strain.
• Figure 5: Data collection in MuJoCo of whiskers sweeping past (A) a power drill, (B) a strawberry, (C) a clamp, and (D) a coffee can in the YCB dataset. Red points indicate potential contact positions on the surface at the whisker sweeping plane.