SuperTrack: Motion Tracking for Physically Simulated Characters using Supervised Learning
Created on Nov. 17, 2021, 1:11 p.m.
This year at SIGGRAPH Asia we will be presenting SuperTrack: Motion Tracking for Physically Simulated Characters using Supervised Learning - a method for tracking motion data using physically simulated characters which relies purely on supervised learning. The trick is to train a world-model to predict the movement of the physically simulated character which acts as a differentiable simulator through which a control policy can be optimized for directly. Compared to previous methods that rely on reinforcement learning we achieve better animation quality, faster training time, and can scale-up to much larger databases of animation.
Webpage • Paper • Video • Article
Abstract: In this paper we show how the task of motion tracking for physically simulated characters can be solved using supervised learning and optimizing a policy directly via back-propagation. To achieve this we make use of a world model trained to approximate a specific subset of the environment's transition function, effectively acting as a differentiable physics simulator through which the policy can be optimized to minimize the tracking error. Compared to popular model-free methods of physically simulated character control which primarily make use of Proximal Policy Optimization (PPO) we find direct optimization of the policy via our approach consistently achieves a higher quality of control in a shorter training time, with a reduced sensitivity to the rate of experience gathering, dataset size, and distribution.
