← ArchivePaper2023
MuscleVAE: Model-Based Controllers of Muscle-Actuated Characters
Integrates fatigue dynamics into Hill-type muscle model and trains a VAE motion controller from large unstructured datasets.
Abstract
In this paper, we present a simulation and control framework for generating biomechanically plausible motion for muscle-actuated characters. We incorporate a fatigue dynamics model, the 3CC-r model, into the widely-adopted Hill-type muscle model to simulate the development and recovery of fatigue in muscles, which creates a natural evolution of motion style caused by the accumulation of fatigue from prolonged activities. To address the challenging problem of controlling a musculoskeletal system with high degrees of freedom, we propose a novel muscle-space control strategy based on PD control. Our simulation and control framework facilitates the training of a generative model for muscle-based motion control, which we refer to as MuscleVAE. By leveraging the variational autoencoders (VAEs), MuscleVAE is capable of learning a rich and flexible latent representation of skills from a large unstructured motion dataset, encoding not only motion features but also muscle control and fatigue properties. We demonstrate that the MuscleVAE model can be efficiently trained using a model-based approach, resulting in the production of high-fidelity motions and enabling a variety of downstream tasks.
How to read this
- Category
- Method: a model-based controller for muscle-actuated characters
- Contributions
- A simulation/control framework integrating the 3CC-r fatigue dynamics model into a Hill-type muscle model to evolve motion style as fatigue accumulates
- A muscle-space control strategy based on PD control to handle the high-DoF musculoskeletal system
- MuscleVAE, a VAE trained model-based from large unstructured motion data, encoding motion plus muscle control and fatigue, enabling high-fidelity motions and downstream tasks
- Context
- Builds on scalable muscle-actuated human simulation and control (Lee et al. 2019), adding fatigue dynamics and a learned generative controller.Builds on: Scalable Muscle-Actuated Human Simulation and Control
- Correctness
- Plausibility rests on the Hill-type-plus-3CC-r muscle/fatigue model and a model-based VAE training scheme; results are demonstrated in simulation, so biomechanical realism is bounded by the muscle model assumptions and the unstructured dataset used.
- Clarity
- Technical but well-motivated; a first pass conveys the fatigue-plus-VAE idea, a second pass is needed for the muscle-space PD control and model-based training.
- How to read it
- First pass for the fatigue-model integration and what MuscleVAE encodes; second pass on the muscle-space control and model-based VAE training if simulating musculoskeletal characters.
Builds on
Built upon by
Nothing yet.
Related work
- ControlVAE: Model-Based Learning of Generative Controllers for Physics-Based Characters 2022 / SIGGRAPH Asia
- Generative GaitNet 2022 / SIGGRAPH
- ReGAIL: Toward Agile Character Control From a Single Reference Motion 2024 / MIG
- MoConVQ: Unified Physics-Based Motion Control via Scalable Discrete Representations 2024 / TOG
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →