Course Recent Advances in Haptic Rendering and Applications

Course: Recent Advances in Haptic Rendering and Applications Session IV: Rendering of Textures and Deformable Surfaces Haptic Rendering of Textured Surfaces Miguel A. Otaduy ETH-Zurich http: //graphics. ethz. ch/~otmiguel otaduy@inf. ethz. ch

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

Introduction

Introduction • Geometric surface texture: – Compelling cue to object identity – Strongly influences forces during manipulation • Objects with rich surface texture information cannot be handled by state-of-the-art haptic rendering methods.

Models Coarse geometric representations Haptic textures

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

3 -Do. F Texture Rendering • 1 contact point on a textured surface – Minsky [1995], Ho et al. [1999]: high frequency forces based on gradient of height field contact point simplified surface height field in texture map

3 -Do. F Texture Rendering • 1 contact point on a textured surface – Siira and Pai [1996]: stochastic model – Pai et al. [2001]: auto-regressive model for roughness and friction

6 -Do. F Texture Rendering • Object-object interaction – Contact cannot be described as point-surface contact – Force and torque output has 6 -Do. F; point contact only has 3 -Do. F • A different rendering algorithm is required

Rendering Algorithm 1) Compute contact information between low-res models

Rendering Algorithm 1) Compute contact information between low-res models 2) Refine contact information using detail geometry stored in textures

Rendering Algorithm 1) Compute contact information between low-res models 2) Refine contact information using detail geometry stored in textures 3) Compute contact forces based on novel texture force model

Force Model Overview • Accounts for important factors identified by perceptual studies • Based on the gradient of inter-object penetration depth • GPU-based computation of directional penetration depth

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

Related Work: Perception & Psychophysics • Studies on perception of textures through a rigid probe by Klatzky and Lederman [1999 present] – Analyze effects of probe diameter, applied force and exploratory speed – Inspiration for our force model

Roughness Vs. Texture Spacing log (roughness) [Klatzky and Lederman 1999 -present] Probe Diameter (D) Applied Force (F) Exploratory Speed (v) log (texture frequency)

Effect of Probe Diameter (D) [Klatzky and Lederman 1999 -present] log (roughness) - Strong influence of geometry D + log (texture frequency)

Effect of Applied Force (F) [Klatzky and Lederman 1999 -present] log (roughness) + F Roughness grows with applied force - log (texture frequency)

Effect of Exploratory Speed (v) log (roughness) [Klatzky and Lederman 1999 -present] Dynamic effects already present in haptic simulation + v - log (texture frequency)

Offset Surfaces Spherical probe trajectory = offset surface Use offset surface as descriptor of vibration Arbitrary objects ? ? ? How can we generalize offset surfaces?

Penetration Depth: Definition = Minimum translational distance to separate two intersecting objects

Penetration Depth: Definition = Minimum translational distance to separate two intersecting objects

Directional PD: Definition = Minimum translation along n to separate two intersecting objects n

Directional PD: Definition = Minimum translation along n to separate two intersecting objects n

Offset Surface and PD Offset surface Textured surface

Offset Surface and PD Offset surface Textured surface

Offset Surface and PD penetration depth

Force Model • Penetration depth: – Applicable to arbitrary object-object interaction – Also used in previous single-point rendering methods • Penalty-based potential field:

Force Model • Determine penetration direction n • Force and Torque = Gradient of energy:

Effect of Geometry • Force and torque proportional to gradient of penetration depth High amplitude texture High derivative of penetration depth Large force/torque Method validated by Minsky [1995]

Effect of Applied Force • Normal force: • Other forces and torques: Larger normal force Larger roughness effect

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

Directional PD: Definition = Minimum translation along n to separate two intersecting objects n

Directional PD of Height Fields B n A

PD with Texture Images High-res Surface Low-res Surface n

PD with Texture Images High-res Surface Low-res Surface n

PD with Texture Images High-res Surface Low-res Surface n

PD with Texture Images High-res Surface Low-res Surface n

PD Computation Algorithm

Low-Resolution Models…

…+ Texture Images

Step 1: Approximate PD

Step 1: Approximate PD

Step 2: Refined PD

Pass 1: Render Geometry

Pass 1: Texture Mapping

Pass 1: Sample Surfaces

Pass 1: Project to PD Direction

Discrete Height Fields

Pass 2: Subtract Height Fields

Pass 2: Copy to Depth Buffer

Binary Search for Max = PD [Govindaraju et al. 2004]

Test

Test

Gradient of PD

Gradient of PD • Central differences • Recompute PD at 2 new object configurations

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

Experiments • Offline analysis of force model • Quality of texture effects • Performance tests.

Offline Experiments u kh n mh bh D δn v

Effect of Probe Diameter Studies by Klatzky and Lederman Simulation results

Effect of Applied Force Studies by Klatzky and Lederman Simulation results

Effect of Exploratory Speed Studies by Klatzky and Lederman Simulation results

Roughness under Translation x z y

Roughness under Rotation n

Complex Objects

Timings: File and CAD Part File full-res: 285 Ktris File low-res: 632 tris CAD full-res: 658 Ktris CAD low-res: 720 tris Dual Pentium 4 2. 4 GHz NVidia FX 5950

Outline • Motivation • Algorithm Overview • Synthesis of the Force Model • Penetration Depth on the GPU • Experiments and Results • Conclusion

Summary • Haptic rendering algorithm using low-res models and texture images • Force model inspired by psychophysics studies • Image-space algorithm for PD computation (implemented on GPU)

Results • 500 Hz force update rate with relatively simple models • 100 Hz-200 Hz force update rate with complex models • Haptic rendering of interaction between complex textured models

Limitations • Cannot handle surfaces that cannot be described as height fields in the contact region • Possible sampling-related aliasing • Limited stability with high PD gradient

Future Work • Higher frequency textures • Deformable textured surfaces • Analysis of human factors

References Haptic Display of Interaction between Textured Models. Miguel A. Otaduy, Nitin Jain, Avneesh Sud and Ming C. Lin. In Proc. of IEEE Visualization Conference 2004. A Perceptually-Inspired Force Model for Haptic Texture Rendering. Miguel A. Otaduy and Ming C. Lin. In Proc. of the Symposium on Applied Perception in Graphics and Visualization 2004. http: //gamma. cs. unc. edu/HTextures/