Simulating Biped Behaviors from Human Motion Data

[source] siggraph

[year] 2007

模仿人的动作。

修改动作数据，PDs驱动

 

Imitation-based learning appoach使可以迁移运动风格，但需要精细的控制。因此有本文的思路。

得到的控制器还可以在几个间转移。

 

相关工作：

1. Data-Driven using Mocap data.  Motion Graph

2. dynamic control systems

3. Dynamic motion synthesis as variational optimization problem

4. ML techniques

 

Data Collection and Processing

 

Motion Rectification

   因为表演人物与模型不一致，且为平面模型

Formulation

  \hat m = m + d

  d = \sum_{j=1}^{m}{h_{ij}B_j(t;c_j,w_j)}

  B = 1/2(1+cos(\pi/w(t-cj)))  c_j – w_j < t < c_j + w_j时

  每个node恰为swing takeoff and kickdown of a stance foot or with the halfway of a swing phase

  因此，要确定  h 及 w

 

  目标：使PDs模拟的动作与原始动作接近

Optimization Method

  各种优化算法的选择  : randomly choose initial parameters and run a downhill simplex method reapeatedly with diff. init. param. to find a local extremum.

Behavior Control

  构造stationary和transition controller以使状态空间可以延续

  方法: regression techniques that selects nearby samples in the state space and combines the output poses at those samples to produce a desired output.

  近邻的计算，提出考量几个要素间距离

 

 

 

个人以为重要的是思路，对动作数据作修改而非制作精细的控制器算法。

转载于:https://www.cnblogs.com/justin_s/archive/2011/06/06/2073689.html