Navigation and 3D Surface Reconstruction from Passive Whisker Sensing

TL;DR

This work presents a passive whisker-based sensing framework for navigation and partial 3D surface reconstruction on a robot arm. By combining 3D contact localization via Bayesian filtering with a calibrated Gaussian Process sensor model, the method achieves sub-millimeter accuracy in locating contact points and leverages contact histories to reconstruct object surfaces. An occupancy-map approach using Bayesian Hilbert Maps enables integration of multiple whiskers and time histories to produce coherent scene representations. The combination of unobtrusive sensing, real-time proximity feedback, and map-based planning has practical implications for safe, information-rich operation in cluttered environments where vision is limited.

Abstract

Whiskers provide a way to sense surfaces in the immediate environment without disturbing it. In this paper we present a method for using highly flexible, curved, passive whiskers mounted along a robot arm to gather sensory data as they brush past objects during normal robot motion. The information is useful both for guiding the robot in cluttered spaces and for reconstructing the exposed faces of objects. Surface reconstruction depends on accurate localization of contact points along each whisker. We present an algorithm based on Bayesian filtering that rapidly converges to within 1\,mm of the actual contact locations. The piecewise-continuous history of contact locations from each whisker allows for accurate reconstruction of curves on object surfaces. Employing multiple whiskers and traces, we are able to produce an occupancy map of proximal objects.

Figures (13)

• Figure 1: A)Instruments robot end-effector with an array of 16 semi-curved whisker sensors. B)Illustrates the sensing region of a semi-curved sensor. Shaded region approximately defines a threshold or keep-out region based on the amount of deflection measured on the sensor. C)Shows the combined sensing region and keep-out region of a sensorized end-effector. Contacts within the threshold generate a repelling force $F_r$.
• Figure 2: Curved whiskers with different geometries and sensing regions (A). Depending on motion direction, a straight whisker may buckle, making interpretation more difficult (B).
• Figure 3: Figure illustrates a whisker sensor brushing along the surface of an object. The trace of the contact locations, $p_c$, over time informs the robot about the object surface.
• Figure 4: Left: 3-axis positioning stage used for calibrating whisker sensors. Right: Example of different calibration points collected within the sensing region.
• Figure 5: Sensor data fitted with Gaussian Process Regression. Top plots use Radial Basis Function kernel. Bottom plots use Thin-plate kernel.