Haptic Shape Display
with Michael Rittikaidachar and Ye Wang. Completed for Stanford’s ME 327: Design & Control of Haptic Systems in Spring 2018.
Overview
Shape displays are a unique class of tangible interfaces that allow the shape of any object or surface to be physically rendered in 2.5D. While these devices have enabled a variety of unique physical interactions with digital media, they have largely prioritized shape output over haptic rendering. In this work, completed for Stanford's ME 327 Design & Control of Haptic Systems, we expand the capabilities of existing shape display hardware towards the goal of enabling richer haptic interaction with shapes rendered by the display. We implement variable stiffness rendering and surface deformations that dynamically respond to a user's touch. We demonstrated this on a shape display of 8x1 non-backdrivable pins running an admittance control algorithm to determine the proper motion of the pins in response the force applied. We detail our methodology and implementation, and present results on a preliminary system evaluation comparing the measured and expected motion of a single pin in multiple loading scenarios.
Challenges & Solutions
In most standard haptics applications, haptic devices use what's known as impedance control. Under impedance control, the system is backdrivable, meaning the user is able to freely move the actuated joints or links when the motors are off. In this way, the system is able to use displacement as an input to determine how much force should be applied, based on what the user should feel. Looking at it from the simple case of Hooke's law, F = kx, impedance control uses the measured x to determine the required F based on a desired stiffness k. This has the benefit of only requiring position, not force sensing, which is often costly and not scalable. Since the pins on our existing system shapeShift is not backdrivable, we couldn't use impedance control.
We use the inverse, admittance control, where we measure the applied force and determine the motion of the pins based on the desired perceived stiffness. Using one force sensor for each pin would be costly and impractical for the current implementation, since the pins are moving and have only 5 mm width. Instead, we developed a load cell platform that measures the net force on the entire row and integrate that information with capacitive touch sensing on each pin to determine which pins are in contact.
Implementation
Dynamics & Control
To enable the dynamic rendering of mechanical properties of any arbitrary shape, we developed a physical model of a shape display. We modeled the display as an array of mass-spring-damper systems, each governed by independent ordinary second-order differential equations. To account for surface-level deformation, an ad hoc distribution factor was used to determine how much force the neighboring pins surrounding the contact point would feel. This allows us to render objects who's surfaces appear both completely rigid or completely uncoupled, independently from the stiffness felt at the contact point.
Based on the desired impedance parameters and measured forces, the desired pin positions were integrated numerically and delivered to the shape display.