# 維持燒杯水平上移
## Ver1
<code>getOrientation()</code>
\begin{equation*}
R_{3,3} =
\begin{pmatrix}
R[0] & R[1] & R[2] \\
R[3] & R[4] & R[5] \\
R[6] & R[7] & R[8] \\
\end{pmatrix}
\end{equation*}
$p' = Rp + T$
Where $p$ is a point whose coordinates are given with respect to the local coordinate system of a node, $R$ the rotation matrix returned by the <code>getOrientation()</code> function, T is the position returned by the <code>getPosition()</code> function and $p'$ represents the same point but this time with coordinates expressed in the global (world) coordinate system.

Reward: 將原本碰點reward設計，加上-(1.0-R[4])
神經網路輸出$3^4=81$種action控制最末四個馬達motor 0~6中的motor 3,4,5,6

上述觀察為末端馬達旋轉後推擠木箱導致第1馬達轉動，又因為神經網路input不包含此馬達之position因此無法調適。

## Ver2
強迫motor 0,1,2維持原本position，透過每次採取action時候也指定遠本position與馬達轉速給motor 0,1,2
:face_with_one_eyebrow_raised: 看起來成功了對ㄇ？
:face_palm: 收斂到變成原地直接不動了，少動少錯不動不錯
(影片待補，網路不穩)

## Ver3
將軒裕詠太構想增加兩個關節自由移動