Spherical Latent Motion Prior for Physics-Based Simulated Humanoid Control
🤖 AI Summary
This work addresses the challenge of generating diverse and physically plausible humanoid motions with existing physics-based control methods—such as VAEs or AMP—which often suffer from information loss or mode collapse. The authors propose a two-stage framework: first, a high-fidelity motion-tracking controller is trained; then, its policy is distilled into a spherical latent action space, augmented with a discriminator and local semantic consistency constraints to structure the latent space. This approach uniquely integrates spherical latent representations, policy distillation, and adversarial discrimination, enabling stable, semantically meaningful random sampling while preserving fine-grained motion details. Evaluated on a newly collected two-hour martial arts motion capture dataset, the method generates realistic, physically valid, and morphology-agnostic combat behaviors using only simple rule-based rewards.
Application Category
📝 Abstract
Learning motion priors for physics-based humanoid control is an active research topic. Existing approaches mainly include variational autoencoders (VAE) and adversarial motion priors (AMP). VAE introduces information loss, and random latent sampling may sometimes produce invalid behaviors. AMP suffers from mode collapse and struggles to capture diverse motion skills. We present the Spherical Latent Motion Prior (SLMP), a two-stage method for learning motion priors. In the first stage, we train a high-quality motion tracking controller. In the second stage, we distill the tracking controller into a spherical latent space. A combination of distillation, a discriminator, and a discriminator-guided local semantic consistency constraint shapes a structured latent action space, allowing stable random sampling without information loss. To evaluate SLMP, we collect a two-hour human combat motion capture dataset and show that SLMP preserves fine motion detail without information loss, and random sampling yields semantically valid and stable behaviors. When applied to a two-agent physics-based combat task, SLMP produces human-like and physically plausible combat behaviors only using simple rule-based rewards. Furthermore, SLMP generalizes across different humanoid robot morphologies, demonstrating its transferability beyond a single simulated avatar.
Problem
Research questions and friction points this paper is trying to address.
motion priors
physics-based humanoid control
information loss
mode collapse
diverse motion skills
Innovation
Methods, ideas, or system contributions that make the work stand out.
Spherical Latent Motion Prior
physics-based humanoid control
motion prior distillation
structured latent space
cross-morphology generalization
🔎 Similar Papers
No similar papers found.
