Method and system for simulating character
AssigneeSnu R&Db Foundation
InventorsJehee Lee, Manmyung Kim, Kwang Won Sok
Life of this patent
Estimating the term…
Abstract
A method and system for simulating a character is provided. The method of simulating a character including: optimizing motion data by using a displacement mapping and a Proportional Derivative (PD) control; and performing controller training by using the optimized motion data and controlling a motion of the character. In this instance, the optimizing includes: generating a target motion by using the displacement mapping between an input motion and a displacement parameter; and generating a simulated motion by using the target motion and an objective function.
Claims28
A method of using a non-transitory computer readable medium for storing instructions which when executed by a computer processor causes the computer to perform a method of simulating a character, the method comprising:
- optimizing recorded motion data captured from human behavior using a displacement mapping and a Proportional Derivative (PD) control;
- and
- performing controller training using the optimized motion data and controlling a motion of the character, wherein
- the performing of the controller training and controlling comprises:
- performing controller training using a collection of state-action trajectories;
- improving the controller by using the motion data and the displacement mapping;
- and
- controlling a transition between states of behaviors by using the controller,
- and wherein the performing of the controller training using the collection of state-action trajectories comprises:
- selecting at least one sample near a novel state when the novel state of the character is provided;
- and
- calculating a target pose by combining at least one output pose included in the at least one sample.
A system for simulating a character, the system comprising:
- a motion data optimization unit optimizing recorded motion data captured from human behavior using a displacement mapping and a PD control;
- a character control unit performing controller training using the optimized motion data and controlling a motion of the character;
- a controller training unit configured to perform controller training by using a collection of state-action trajectories;
- a controller improvement unit configured to improve the controller by using the motion data and the displacement mapping;
- and
- a transition control unit configured to control a transition between states of behaviors by using the controller,
- wherein the controller training unit selects at least one sample near a novel state when the novel state of the character is provided at a subsequent time instance, and
- calculates a target pose by combining at least one output pose included in the at least one sample.
Dependencies are inferred from the claim text.