SMP: Reusable Score-Matching Motion Priors for Physics-Based Character Control
This work addresses the need for reusable motion priors in physics-based character control, enabling more efficient and modular training of virtual agents without retraining for each new task.
The paper tackles the problem of limited reusability in adversarial imitation learning for motion priors by introducing Score-Matching Motion Priors (SMP), which uses pre-trained diffusion models and score distillation to create task-agnostic priors that can be frozen and reused across various control tasks, achieving high-quality motion comparable to state-of-the-art methods.
Data-driven motion priors that can guide agents toward producing naturalistic behaviors play a pivotal role in creating life-like virtual characters. Adversarial imitation learning has been a highly effective method for learning motion priors from reference motion data. However, adversarial priors, with few exceptions, need to be retrained for each new controller, thereby limiting their reusability and necessitating the retention of the reference motion data when training on downstream tasks. In this work, we present Score-Matching Motion Priors (SMP), which leverages pre-trained motion diffusion models and score distillation sampling (SDS) to create reusable task-agnostic motion priors. SMPs can be pre-trained on a motion dataset, independent of any control policy or task. Once trained, SMPs can be kept frozen and reused as general-purpose reward functions to train policies to produce naturalistic behaviors for downstream tasks. We show that a general motion prior trained on large-scale datasets can be repurposed into a variety of style-specific priors. Furthermore SMP can compose different styles to synthesize new styles not present in the original dataset. Our method produces high-quality motion comparable to state-of-the-art adversarial imitation learning methods through reusable and modular motion priors. We demonstrate the effectiveness of SMP across a diverse suite of control tasks with physically simulated humanoid characters. Video demo available at https://youtu.be/ravlZJteS20