Biomimetic whisker transducer for vibrissal tactile sensing

Abstract

The unique vibrissal tactile perception possessed by rats allows them to explore the surrounding environment actively with their whiskers, and helps them survive and thrive in hostile environments like underground tunnels, where vision becomes of little use. On the other hand, today most robots lack tactile sensing ability and rely on machine vision heavily. This seriously limits their survival chance and exploration performance in unstructured environments. In this thesis, a novel whisker transducer (WT) aiming to mimic the biological vibrissa system has been proposed, which can endow robots with vibrissal tactile sensing ability. The WT was developed based on a novel transduction matrix model, which characterizes the forward actuation and backward sensing functions of the transducer with a 2×2 matrix. As a result, the WT can perform active whisking and simultaneously sense the mechanical stimuli at its tip, just like what rat's whisker does. And the force, velocity and mechanical impedance at whisker tip are sensed from the input electrical signals, without using any additional sensors. The concept and functions of the WT were validated numerically using a finite element model and experimentally using a prototype. It can accurately measure mechanical impedance, force and velocity at whisker tip with errors of 1.28%, 0.88% and 1.55%, respectively. It also possesses features such as compact structure, good versatility and easy instrumentation, which facilitate its deployment on robots. Vibrissal tactile texture sensing and object locating were first investigated. The texture sensing experiment showed that when the WT sweeps across a surface, the force and velocity sensed at whisker tip both have strong linear relationship with roughness (regression analysis R 2 > 0.99). And the damping property of the surface can be obtained from real part of WT-measured mechanical impedance. The WT was also used to study the mechanism of rat's vibrissal tactile texture perception in a biomimetic perspective. Then the object locating experiment was investigated with a miniaturized WT (50×50×5000 µm). A bio-inspired spatial-temporal-intensity triple-coding scheme was proposed and implemented using digital Abstract ii signal processing and artificial neural networks. The WT can detect horizontal and radial locations of a contacting object with measurement uncertainties of 0.27 µm and 5.4 µm, respectively. And a WT array can enable object locating in the vertical axis. Potential applications of WT were explored experimentally. First, the miniaturized WT can accurately measure geometry of high-aspect-ratio micro trenches and detect invalid tip contact caused by protrusion on the sidewall. Second, a modified WT with a sphere tip can sense both viscosity and density of a fluid online. The extracted viscosity and density were separated from each other, rather than in the form of product (ηρ) or quotient (η/ρ) as detected by existing viscosity sensors. The low working frequency (104 Hz) and large vibration amplitude (0.43 mm) of WT made ...