Animating human dressing

Alexander Clegg, Jie Tan, Greg Turk, C. Karen Liu

TL;DR

This paper introduces a novel system for animating human characters dressing themselves, a complex task that has been largely absent from computer animation. By breaking down dressing into a set of primitive actions and using a cloth-aware feedback controller, the system can generate physically plausible and stylistically diverse dressing animations.

What Problem Does This Paper Address?

Animating human dressing is surprisingly challenging. While a common daily activity, it requires intricate coordination between a deformable garment and a complex articulated human body. Existing animation techniques struggle with this due to the dynamic nature of cloth, the need for precise body-garment interaction, and the difficulty of planning motions that account for both. This paper aims to bridge this gap by providing a systematic approach to synthesizing realistic human dressing animations, allowing animators to control the process with high fidelity.

Methods

The core of the system lies in an "action-based" approach. The authors first identify a set of "primitive actions" that constitute the majority of human dressing motions, such as aligning a hand with an armhole or pulling a garment onto a limb. An animator assembles these primitive actions into a sequence to define a desired dressing motion.

The system then employs a "dressing controller" that handles each primitive action. This controller uses a combination of feed-forward and feedback mechanisms. For "alignment" actions, where precise interaction with the cloth is critical, a cloth-aware feedback controller is used. This controller plans a path for the character's end-effector (e.g., hand) towards a target cloth feature (e.g., an armhole) by iteratively finding intermediate goals based on geodesic distances on the cloth. This approach accounts for the dynamic and often occluded nature of cloth features.

For "traversal" actions, where the garment is being pulled onto the body, a feed-forward controller is used. This controller plans the entire joint trajectory for the action at the beginning, often using inverse kinematics (IK) and collision-free path planning (RRT) to ensure smooth and collision-free motion. Other primitive actions like "tracking" (following a reference motion), "grip," "release," and "idling" are also incorporated to complete the dressing sequence. Crucially, the system uses a physics-based cloth simulator (ARCSim) to drive the garment's motion, and the character's motion is adapted to interact naturally with the simulated cloth.

Datasets

Not discussed in the paper. The authors mention using garments from the "Berkeley Garment Library" and editing them to fit their character. Motion capture data and sparse keyframes are used as reference motions, but these are not presented as formal datasets.

Key Insights and Ideas

  • Decomposition into Primitive Actions: The insight that a vast majority of dressing motions can be broken down into a small set of primitive actions is key to making the problem tractable. This allows for a modular and controllable animation system.
  • Cloth-Aware Feedback Control for Alignment: Recognizing that precise interaction with dynamic cloth is crucial for certain actions (like entering a sleeve) and that traditional path planning is insufficient, the authors developed a feedback controller that actively adapts to the cloth's state.
  • Geodesic Distances for Cloth Feature Targeting: Using geodesic distances on the cloth surface to find intermediate goals for alignment is an elegant way to navigate complex cloth shapes and occlusions without full online cloth simulation.
  • Robustness to Cloth Resolution: The system's ability to adapt to different cloth mesh resolutions without manual tuning demonstrates the robustness of the control strategy.

Key Results

The system successfully animated characters dressing in various garments, including jackets, shorts, robes, and vests, using different styles. The authors demonstrated:

  • Jacket Dressing: The system successfully animated putting on a jacket, including complex maneuvers like swinging an arm to the back and aligning with an occluded armhole.
  • Shorts Dressing: Animations for putting on shorts in both sitting and standing poses were generated, showcasing the system's versatility.
  • Vest Dressing: A dynamic vest-dressing sequence demonstrated the system's ability to handle fast-moving cloth features.
  • Assisted Dressing: The system was extended to simulate a scene where one character assists another in putting on a robe, highlighting its potential for collaborative animation.
  • Robustness: The system demonstrated robustness to changes in cloth mesh resolution, adapting the animation without manual intervention.

The typical running time for a dressing animation was several hours to a day, with cloth simulation being the most time-consuming part.

Limitations and Future Work

The authors acknowledge several limitations:

  • Tactile Feedback: The current system relies on visual feedback for cloth interaction. Real humans often use tactile feedback, which is not simulated here.
  • One-Way Coupling: The cloth simulation is one-way coupled, meaning the cloth does not exert forces on the human character. This can lead to unrealistic cloth behavior (e.g., exceeding strain limits) and physically impossible human motions (e.g., lack of balance control).
  • Short-Sightedness: The current algorithm is "shortsighted" and relies on user input for complex long-term planning motions like re-gripping or catching a dropped garment.
  • Initial Cloth Configuration: The system assumes a favorable initial cloth configuration. Tangled or improperly positioned garments would likely cause the system to fail.
  • Sequential Actions and Failure Handling: Actions are executed sequentially, and the system does not automatically respond to failure scenarios. It also requires the user to specify the action sequence.

Future work suggested includes:

  • Dexterous Manipulation: Incorporating more sophisticated hand-object interaction, such as gripping and fastening with buttons or zippers.
  • Two-Way Coupling: Implementing two-way coupling between cloth and human dynamics to improve realism and balance control.
  • Long-Term Planning: Developing methods for incorporating approximate cloth simulation to enable better long-term planning.
  • Automated Strategy Generation: Creating a system that can automatically determine a high-level dressing strategy for a new garment.
  • Multi-Layer Cloth Simulation: Improving cloth simulation to handle multiple layers of clothing effectively.
  • Tight Contact Handling: Enhancing the simulation of tight contact between cloth and the human figure.

Summary updated

Abstract

Dressing is one of the most common activities in human society. Perfecting the skill of dressing can take an average child three to four years of daily practice. The challenge is primarily due to the combined difficulty of coordinating different body parts and manipulating soft and deformable objects (clothes). We present a technique to synthesize human dressing by controlling a human character to put on an article of simulated clothing. We identify a set of primitive actions which account for the vast majority of motions observed in human dressing. These primitive actions can be assembled into a variety of motion sequences for dressing different garments with different styles. Exploiting both feed-forward and feedback control mechanisms, we develop a dressing controller to handle each of the primitive actions. The controller plans a path to achieve the action goal while making constant adjustments locally based on the current state of the simulated cloth when necessary. We demonstrate that our framework is versatile and able to animate dressing with different clothing types including a jacket, a pair of shorts, a robe, and a vest. Our controller is also robust to different cloth mesh resolutions which can cause the cloth simulator to generate significantly different cloth motions. In addition, we show that the same controller can be extended to assistive dressing.
BibTeX
@article{clegg2015animating,
  author = {Alexander Clegg and Jie Tan and Greg Turk and C. Karen Liu},
  title = {{Animating human dressing}},
  journal = {ACM Trans. Graphics},
  year = {2015},
  doi = {10.1145/2766986},
}

Loading related content...

Figure