在阅读《Indirect Kalman Filter for 3D Attitude Estimation》一文时,发现文中对四元素的定义采用了非Hamilton方式,阅读有一些冲突,难以转换,至此,决定将该文章转换为Hamilton的表达形式。
1 四元素定义:
q
=
q
w
+
q
x
i
+
q
y
j
+
q
z
k
(
1
)
\bm{q} = q_w + q_x \bm{i} + q_y\bm{j} + q_z\bm{k} \qquad (1)
q=qw+qxi+qyj+qzk(1)
其中,虚数单位:
i
\bm{i}
i,
j
\bm{j}
j,
k
\bm{k}
k满足如下关系:
i
2
=
j
2
=
k
2
=
i
j
k
=
−
1
(
2
)
\bm{i}^2 = \bm{j}^2 = \bm{k}^2 = \bm{ijk} = -1\qquad(2)
i2=j2=k2=ijk=−1(2)
从而:
i
j
=
−
j
i
=
k
,
j
k
=
−
k
j
=
i
,
k
i
=
−
i
k
=
j
(
3
)
\bm{ij} = -\bm{ji} = \bm{k},\quad \bm{jk} = -\bm{kj} = \bm{i}, \quad \bm{ki} = -\bm{ik} = \bm{j} \qquad (3)
ij=−ji=k,jk=−kj=i,ki=−ik=j(3)
此外,四元数还可以表达成如下式所示的形式:
常量+向量形式:
q
=
q
w
+
q
v
(
4
)
\bm{q} = q_w + \bm{q}_v \qquad(4)
q=qw+qv(4)
其中:
q
v
=
q
x
i
+
q
y
j
+
q
z
k
\bm{q}_v = q_x \bm{i} + q_y\bm{j} + q_z\bm{k}
qv=qxi+qyj+qzk
向量形式
q
=
[
q
w
q
v
]
=
[
q
w
q
x
q
y
q
z
]
(
5
)
\bm{q} = \begin{bmatrix} q_w\\ \bm{q}_v \end{bmatrix} = \begin{bmatrix}q_w\\ q_x \\ q_y \\ q_z \end{bmatrix} \qquad(5)
q=[qwqv]=⎣⎢⎢⎡qwqxqyqz⎦⎥⎥⎤(5)
2 四元素的主要性质
2.1 加法运算
q
1
±
q
2
=
[
q
w
1
q
v
1
]
±
[
q
w
2
q
v
2
]
=
[
q
w
1
±
q
w
2
q
v
1
±
q
v
2
]
(
6
)
\bm{q}_1 \pm \bm{q}_2 = \begin{bmatrix}q_{w1} \\ \bm{q}_{v1} \end{bmatrix} \pm \begin{bmatrix}q_{w2} \\ \bm{q}_{v2} \end{bmatrix} = \begin{bmatrix}q_{w1} \pm q_{w2} \\ \bm{q}_{v1} \pm \bm{q}_{v2} \end{bmatrix} \qquad (6)
q1±q2=[qw1qv1]±[qw2qv2]=[qw1±qw2qv1±qv2](6)
加法运算满足交换律和结合律:
q
1
+
q
2
=
q
2
+
q
1
(
7
)
\bm{q}_1 + \bm{q}_2 = \bm{q}_2 + \bm{q}_1 \qquad (7)
q1+q2=q2+q1(7)
q
1
+
(
q
2
+
q
3
)
=
(
q
1
+
q
2
)
+
q
3
(
8
)
\bm{q}_1 + (\bm{q}_2 + \bm{q}_3) = (\bm{q}_1 + \bm{q}_2) + \bm{q}_3 \qquad(8)
q1+(q2+q3)=(q1+q2)+q3(8)
2.2 四元素数乘
k
q
=
[
k
q
w
k
q
v
]
=
[
k
q
w
k
q
x
k
q
y
k
q
z
]
(
9
)
k\bm{q} = \begin{bmatrix} kq_w \\ k\bm{q}_v \end{bmatrix} = \begin{bmatrix} kq_w\\ kq_x\\ kq_y\\ kq_z \end{bmatrix} \qquad (9)
kq=[kqwkqv]=⎣⎢⎢⎡kqwkqxkqykqz⎦⎥⎥⎤(9)
2.3 四元素点乘
点乘指的是两个四元素每个对应位置上的数值相乘再求和,即:
q
⋅
p
=
q
w
p
w
+
q
x
p
x
+
q
y
p
y
+
q
z
p
z
(
10
)
\bm{q} \cdot \bm{p} = q_w p_w + q_xp_x + q_yp_y + q_zp_z \qquad(10)
q⋅p=qwpw+qxpx+qypy+qzpz(10)
2.4 乘法运算
⊗
\otimes
⊗
p
⊗
q
=
[
p
w
q
w
−
p
x
q
x
−
p
y
q
y
−
p
z
q
z
p
w
q
x
+
p
x
q
w
+
p
y
q
z
−
p
z
q
y
p
w
q
y
−
p
x
q
z
+
p
y
q
w
+
p
z
q
x
p
w
q
z
+
p
x
q
y
−
p
y
q
x
+
p
z
q
w
]
=
[
p
w
q
w
−
p
v
⊤
q
v
p
w
q
v
+
q
w
p
v
+
p
v
×
q
v
]
(
11
)
\bm{p} \otimes \bm{q} = \begin{bmatrix} p_w q_w - p_x q_x - p_y q_y - p_z q_z \\ p_w q_x + p_x q_w + p_y q_z - p_z q_y \\ p_w q_y - p_x q_z + p_y q_w + p_z q_x \\ p_w q_z + p_xq_y - p_y q_x + p_z q_w \end{bmatrix} = \begin{bmatrix} p_w q_w - \bm{p}_v ^\top \bm{q}_v\\ p_w \bm{q}_v + q_w \bm{p}_v + \bm{p}_v \times\bm{q}_v \end{bmatrix} \quad (11)
p⊗q=⎣⎢⎢⎡pwqw−pxqx−pyqy−pzqzpwqx+pxqw+pyqz−pzqypwqy−pxqz+pyqw+pzqxpwqz+pxqy−pyqx+pzqw⎦⎥⎥⎤=[pwqw−pv⊤qvpwqv+qwpv+pv×qv](11)
因(11)式中含有叉乘项,故四元素乘法不满足交换律,即:
p
⊗
q
≠
q
⊗
p
(
12
)
\bm{p} \otimes \bm{q} \neq \bm{q} \otimes \bm{p} \qquad (12)
p⊗q=q⊗p(12)
例外:当
p
v
×
q
v
=
0
\bm{p}_v \times \bm{q}_v = 0
pv×qv=0时,叉乘项为零时满足交换律,此时,
p
v
=
0
\bm{p}_v=0
pv=0 或
q
v
=
0
\bm{q}_v=0
qv=0 或
p
v
∣
∣
q
v
\bm{p}_v||\bm{q}_v
pv∣∣qv(即两个向量平行)。
结合律:
(
q
1
⊗
q
2
)
⊗
q
3
=
q
1
⊗
(
q
2
⊗
q
3
)
(
13
)
(\bm{q}_1 \otimes \bm{q}_2) \otimes \bm{q}_3 = \bm{q}_1 \otimes (\bm{q}_2 \otimes \bm{q}_3) \qquad(13)
(q1⊗q2)⊗q3=q1⊗(q2⊗q3)(13)
分配律:
q
1
⊗
(
q
2
+
q
3
)
=
q
1
⊗
q
2
+
q
1
⊗
q
3
(
14
)
\bm{q}_1 \otimes (\bm{q}_2 + \bm{q}_3) = \bm{q}_1 \otimes \bm{q}_2 + \bm{q}_1 \otimes \bm{q}_3 \qquad(14)
q1⊗(q2+q3)=q1⊗q2+q1⊗q3(14)
(
q
1
+
q
2
)
⊗
q
3
=
q
1
⊗
q
3
+
q
2
⊗
q
3
(
15
)
(\bm{q}_1 + \bm{q}_2) \otimes \bm{q}_3 = \bm{q}_1 \otimes \bm{q}_3 + \bm{q}_2 \otimes \bm{q}_3 \qquad(15)
(q1+q2)⊗q3=q1⊗q3+q2⊗q3(15)
矩阵乘法形式:
根据(11)式,提取出相应的四元素向量,便可以得到:
q
1
⊗
q
2
=
[
q
1
]
L
q
2
=
[
q
w
1
−
q
x
1
−
q
y
1
−
q
z
1
q
x
1
q
w
1
−
q
z
1
q
y
1
q
y
1
q
z
1
q
w
1
−
q
x
1
q
z
1
−
q
y
1
q
x
1
q
w
1
]
q
2
(
16
)
\bm{q}_1 \otimes \bm{q}_2 = [\bm{q}_1]_L \bm{q}_2=\begin{bmatrix} q_{w1} & -q_{x1} & -q_{y1} & -q_{z1} \\ q_{x1} & q_{w1} & -q_{z1} & q_{y1} \\ q_{y1} & q_{z1} & q_{w1} & -q_{x1} \\ q_{z1} & -q_{y1} & q_{x1} & q_{w1} \end{bmatrix} \bm{q}_2 \qquad (16)
q1⊗q2=[q1]Lq2=⎣⎢⎢⎡qw1qx1qy1qz1−qx1qw1qz1−qy1−qy1−qz1qw1qx1−qz1qy1−qx1qw1⎦⎥⎥⎤q2(16)
q
1
⊗
q
2
=
[
q
2
]
R
q
1
=
[
q
w
2
−
q
x
2
−
q
y
2
−
q
z
2
q
x
2
q
w
2
q
z
2
−
q
y
2
q
y
2
−
q
z
2
q
w
2
q
x
2
q
z
2
q
y
2
−
q
x
2
q
w
2
]
q
1
(
17
)
\bm{q}_1 \otimes \bm{q}_2 = [q_2]_R \bm{q}_1 = \begin{bmatrix} q_{w2} & -q_{x2} & -q_{y2} & -q_{z2} \\ q_{x2} & q_{w2} & q_{z2} & -q_{y2} \\ q_{y2} & -q_{z2} & q_{w2} & q_{x2} \\ q_{z2} & q_{y2} & -q_{x2} & q_{w2} \end{bmatrix} \bm{q}_1 \qquad (17)
q1⊗q2=[q2]Rq1=⎣⎢⎢⎡qw2qx2qy2qz2−qx2qw2−qz2qy2−qy2qz2qw2−qx2−qz2−qy2qx2qw2⎦⎥⎥⎤q1(17)
根据(16)和(17)可以得到:
[
q
]
L
=
[
q
w
−
q
x
−
q
y
−
q
z
q
x
q
w
−
q
z
q
y
q
y
q
z
q
w
−
q
x
q
z
−
q
y
q
x
q
w
]
,
[
q
]
R
=
[
q
w
−
q
x
−
q
y
−
q
z
q
x
q
w
q
z
−
q
y
q
y
−
q
z
q
w
q
x
q
z
q
y
−
q
x
q
w
]
q
1
(
18
)
[\bm{q}]_L = \begin{bmatrix} q_{w} & -q_{x} & -q_{y} & -q_{z} \\ q_{x} & q_{w} & -q_{z} & q_{y} \\ q_{y} & q_{z} & q_{w} & -q_{x} \\ q_{z} & -q_{y} & q_{x} & q_{w} \end{bmatrix}, \quad [\bm{q}]_R = \begin{bmatrix} q_{w} & -q_{x} & -q_{y} & -q_{z} \\ q_{x} & q_{w} & q_{z} & -q_{y} \\ q_{y} & -q_{z} & q_{w} & q_{x} \\ q_{z} & q_{y} & -q_{x} & q_{w} \end{bmatrix} \bm{q}_1 \qquad (18)
[q]L=⎣⎢⎢⎡qwqxqyqz−qxqwqz−qy−qy−qzqwqx−qzqy−qxqw⎦⎥⎥⎤,[q]R=⎣⎢⎢⎡qwqxqyqz−qxqw−qzqy−qyqzqw−qx−qz−qyqxqw⎦⎥⎥⎤q1(18)
进一步可以表示为:
[
q
]
L
=
q
w
I
+
[
0
−
q
v
⊤
q
v
q
v
∧
]
,
[
q
]
R
=
q
w
I
+
[
0
−
q
v
⊤
q
v
−
q
v
∧
]
(
19
)
[\bm{q}]_L = q_w \mathbf{I} + \begin{bmatrix} 0 & -\bm{q}_v^\top \\ \bm{q}_v & \bm{q}_v^\land \end{bmatrix}, \qquad [\bm{q}]_R = q_w \mathbf{I} + \begin{bmatrix} 0 & -\bm{q}_v^\top \\ \bm{q}_v & -\bm{q}_v^\land \end{bmatrix} \qquad (19)
[q]L=qwI+[0qv−qv⊤qv∧],[q]R=qwI+[0qv−qv⊤−qv∧](19)
其中:
q
∧
=
[
0
−
q
z
q
y
q
z
0
−
q
x
−
q
y
q
x
0
]
(
20
)
\bm{q} ^ \land = \begin{bmatrix} 0 & -q_z & q_y \\ q_z & 0 & -q_x \\ -q_y & q_x & 0 \end{bmatrix} \qquad (20)
q∧=⎣⎡0qz−qy−qz0qxqy−qx0⎦⎤(20)
是反对称矩阵,即
q
∧
=
−
[
q
∧
]
⊤
\bm{q}^\land = -[\bm{q}^\land]^\top
q∧=−[q∧]⊤,可用于向量间的叉乘运算:
a
×
b
=
a
∧
b
,
∀
a
,
b
∈
R
3
(
21
)
\bm{a} \times \bm{b} = \bm{a}^\land\bm{b}, \quad \forall \bm{a}, \bm{b} \in \mathbb{R}^3 \qquad (21)
a×b=a∧b,∀a,b∈R3(21)
因为:
(
q
⊗
x
)
⊗
p
=
[
p
]
R
[
q
]
L
x
a
n
d
q
⊗
(
x
⊗
p
)
=
[
q
]
L
[
p
]
R
x
(
22
)
(\bm{q} \otimes \bm{x}) \otimes \bm{p} = [\bm{p}]_R[\bm{q}]_L\bm{x} \quad and \quad \bm{q} \otimes(\bm{x} \otimes \bm{p})=[\bm{q}]_L[\bm{p}]_R\bm{x} \qquad(22)
(q⊗x)⊗p=[p]R[q]Lxandq⊗(x⊗p)=[q]L[p]Rx(22)
从而,根据四元素乘法的结合律得:
[
p
]
R
[
q
]
L
=
[
q
]
L
[
p
]
R
(
23
)
[\bm{p}]_R[\bm{q}]_L = [\bm{q}]_L[\bm{p}]_R \qquad (23)
[p]R[q]L=[q]L[p]R(23)
简化写法:
[
q
]
L
=
[
q
Ψ
(
q
)
]
(
24
)
[\bm{q}]_L = [\bm{q} \quad \mathbf{\Psi}(\bm{q})] \qquad(24)
[q]L=[qΨ(q)](24)
其中:
Ψ
(
q
)
=
[
−
q
v
⊤
q
w
I
3
×
3
+
q
v
∧
]
(
25
)
\mathbf{\Psi}(\bm{q}) = \begin{bmatrix} -\bm{q}_v^\top \\ q_w\mathbf{I}_{3\times3} + \bm{q}_v^\land \end{bmatrix} \qquad (25)
Ψ(q)=[−qv⊤qwI3×3+qv∧](25)
同理:
[
q
]
R
=
[
q
Ξ
(
q
)
]
(
26
)
[\bm{q}]_R = [\bm{q} \quad \mathbf{\Xi}(\bm{q})] \qquad(26)
[q]R=[qΞ(q)](26)
其中:
Ξ
(
q
)
=
[
−
q
v
⊤
q
w
I
3
×
3
−
q
v
∧
]
(
27
)
\mathbf{\Xi}(\bm{q}) = \begin{bmatrix} -\bm{q}_v^\top \\ q_w\mathbf{I}_{3\times3} - \bm{q}_v^\land \end{bmatrix} \qquad (27)
Ξ(q)=[−qv⊤qwI3×3−qv∧](27)
此外,对于unit quaternion
q
\bm{q}
q有:
[
q
−
1
]
L
=
[
q
]
L
⊤
(
28
)
[\bm{q}^{-1}]_L = [\bm{q}]_L^\top \qquad (28)
[q−1]L=[q]L⊤(28)
[
q
−
1
]
R
=
[
q
]
R
⊤
(
29
)
[\bm{q}^{-1}]_R = [\bm{q}]_R^\top \qquad (29)
[q−1]R=[q]R⊤(29)
2.5 Identity quaternion
q
1
=
1
=
[
1
0
v
]
(
30
)
\bm{q}_1 = 1 = \begin{bmatrix} 1 \\ \bm{0}_v \end{bmatrix} \qquad (30)
q1=1=[10v](30)
满足:
q
1
⊗
q
=
q
⊗
q
1
=
q
(
31
)
\bm{q}_1 \otimes \bm{q} = \bm{q} \otimes \bm{q}_1 = \bm{q} \qquad(31)
q1⊗q=q⊗q1=q(31)
2.6 共轭
四元素的共轭为:
q
∗
=
[
q
w
−
q
v
]
(
32
)
\bm{q}^* = \begin{bmatrix} q_w\\ -\bm{q}_v \end{bmatrix} \qquad(32)
q∗=[qw−qv](32)
共轭四元素的性质:
q
⊗
q
∗
=
q
∗
⊗
q
=
[
q
w
2
+
q
x
2
+
q
y
2
+
q
z
2
0
v
]
(
32
)
\bm{q} \otimes \bm{q}^* = \bm{q}^* \otimes \bm{q} = \begin{bmatrix} q_w^2 + q_x^2 + q_y^2 + q_z^2 \\ \bm{0}_v \end{bmatrix} \qquad (32)
q⊗q∗=q∗⊗q=[qw2+qx2+qy2+qz20v](32)
以及:
(
p
⊗
q
)
∗
=
q
∗
⊗
p
∗
(
33
)
(\bm{p} \otimes \bm{q})^* = \bm{q}^* \otimes \bm{p}^* \qquad (33)
(p⊗q)∗=q∗⊗p∗(33)
2.7 四元素的模
定义:
∣
∣
q
∣
∣
=
q
⊗
q
∗
=
q
∗
⊗
q
=
q
w
2
+
q
x
2
+
q
y
2
+
q
z
2
(
34
)
||\bm{q}|| = \sqrt{\bm{q} \otimes \bm{q}^*} = \sqrt{\bm{q}^* \otimes \bm{q}} = \sqrt{q_w^2 + q_x^2 + q_y^2 + q_z^2} \qquad (34)
∣∣q∣∣=q⊗q∗=q∗⊗q=qw2+qx2+qy2+qz2(34)
模的性质:
∣
∣
p
⊗
q
∣
∣
=
∣
∣
q
⊗
p
∣
∣
=
∣
∣
p
∣
∣
∣
∣
q
∣
∣
(
35
)
||\bm{p} \otimes \bm{q}|| = ||\bm{q} \otimes \bm{p}|| = ||\bm{p}|| ||\bm{q}|| \qquad (35)
∣∣p⊗q∣∣=∣∣q⊗p∣∣=∣∣p∣∣∣∣q∣∣(35)
简单推导:
∣
∣
p
⊗
q
∣
∣
2
=
(
p
⊗
q
)
⊗
(
p
⊗
q
)
∗
=
p
⊗
q
⊗
q
∗
⊗
p
∗
=
p
⊗
(
q
⊗
q
∗
)
⊗
p
∗
=
p
⊗
[
∣
∣
q
∣
∣
2
0
v
]
⊗
p
∗
=
p
⊗
p
∗
⊗
[
∣
∣
q
∣
∣
2
0
v
]
=
(
∣
∣
p
∣
∣
∣
∣
q
∣
∣
)
2
||\bm{p} \otimes \bm{q}||^2 = (\bm{p} \otimes \bm{q}) \otimes (\bm{p} \otimes\bm{q})^* = \bm{p} \otimes\bm{q}\otimes\bm{q}^*\otimes\bm{p}^* = \bm{p}\otimes(\bm{q}\otimes\bm{q}^*)\otimes\bm{p}^*=\bm{p} \otimes\begin{bmatrix} ||\bm{q}||^2 \\ \bm{0}_v \end{bmatrix} \otimes\bm{p}^* = \bm{p} \otimes \bm{p}^* \otimes \begin{bmatrix} ||\bm{q}||^2 \\ \bm{0}_v \end{bmatrix} = (||\bm{p}||||\bm{q}||)^2
∣∣p⊗q∣∣2=(p⊗q)⊗(p⊗q)∗=p⊗q⊗q∗⊗p∗=p⊗(q⊗q∗)⊗p∗=p⊗[∣∣q∣∣20v]⊗p∗=p⊗p∗⊗[∣∣q∣∣20v]=(∣∣p∣∣∣∣q∣∣)2
2.8 四元素的逆
四元素的逆记为
q
−
1
\bm{q}^{-1}
q−1,,其满足如下条件:
q
⊗
q
−
1
=
q
−
1
⊗
q
=
q
1
(
36
)
\bm{q} \otimes \bm{q}^{-1} = \bm{q}^{-1} \otimes \bm{q} = \bm{q}_1 \qquad(36)
q⊗q−1=q−1⊗q=q1(36)
其中
q
1
\bm{q}_1
q1为"Identity quaternion“, 逆通过下式计算:
q
−
1
=
q
∗
/
∣
∣
q
∣
∣
2
(
37
)
\bm{q}^{-1} = \bm{q}^*/||\bm{q}||^2 \qquad (37)
q−1=q∗/∣∣q∣∣2(37)
2.9 Unit or normalized quaternion
对于unit quaternion有,
∣
∣
q
∣
∣
=
1
||\bm{q}|| = 1
∣∣q∣∣=1,所以:
q
−
1
=
q
∗
(
38
)
\bm{q}^{-1} = \bm{q}^* \qquad(38)
q−1=q∗(38)
3 旋转四元素
形式:
q
=
[
cos
(
θ
/
2
)
u
sin
(
θ
/
2
)
]
=
[
cos
(
θ
/
2
)
u
x
sin
(
θ
/
2
)
u
y
sin
(
θ
/
2
)
u
z
sin
(
θ
/
2
)
]
(
39
)
\bm{q} = \begin{bmatrix} \cos (\theta / 2) \\ \bm{u} \sin (\theta/2) \end{bmatrix} = \begin{bmatrix} \cos (\theta/2) \\ u_x \sin (\theta/2) \\ u_y \sin (\theta/2) \\ u_z \sin (\theta/2) \end{bmatrix} \qquad (39)
q=[cos(θ/2)usin(θ/2)]=⎣⎢⎢⎡cos(θ/2)uxsin(θ/2)uysin(θ/2)uzsin(θ/2)⎦⎥⎥⎤(39)
其中,旋转轴为
u
\bm{u}
u,为单位向量,旋转角度为
θ
\theta
θ,将向量
x
\bm{x}
x绕
u
\bm{u}
u轴旋转
θ
\theta
θ即可表示为:
x
′
=
q
⊗
x
⊗
q
∗
(
40
)
\bm{x}' = \bm{q} \otimes \bm{x} \otimes \bm{q}^* \qquad (40)
x′=q⊗x⊗q∗(40)
可以验证,旋转四元素为unit quaternion,跟unit quaternion一样 具有如下性质:
(
p
⊗
q
)
−
1
=
q
−
1
⊗
p
−
1
(
41
)
(\bm{p} \otimes \bm{q})^{-1} = \bm{q}^{-1}\otimes\bm{p}^{-1} \qquad(41)
(p⊗q)−1=q−1⊗p−1(41)
在介绍四元素基本的定义及相关性质之后,将对《Indirect Kalman Filter for 3D Attitude Estimation》一文当中的相关内容进行转义。
》1.3 Userful Identites
》1.3.1 叉乘反对称矩阵的相关属性
Anti-Commutativity
w
∧
=
−
[
w
∧
]
⊤
(
42
)
\bm{w}^\land = -[\bm{w}^\land]^\top \qquad(42)
w∧=−[w∧]⊤(42)
a
∧
b
=
−
b
∧
a
⇔
−
b
⊤
a
∧
=
a
⊤
b
∧
(
43
)
\bm{a}^\land\bm{b} = -\bm{b}^\land\bm{a} \quad \Leftrightarrow \quad -\bm{b}^\top \bm{a}^\land = \bm{a}^\top\bm{b}^\land \qquad (43)
a∧b=−b∧a⇔−b⊤a∧=a⊤b∧(43)
Distributivity orver Addition
a
∧
+
b
∧
=
(
a
+
b
)
∧
(
44
)
\bm{a}^\land + \bm{b}^\land = (\bm{a} + \bm{b})^\land \qquad(44)
a∧+b∧=(a+b)∧(44)
数乘
c
⋅
w
∧
=
(
c
w
)
∧
(
45
)
c \cdot \bm{w}^\land = (c\bm{w})^\land \qquad(45)
c⋅w∧=(cw)∧(45)
平行向量的叉乘
w
×
(
c
w
)
=
c
⋅
w
∧
w
=
−
c
⋅
(
w
⊤
w
∧
)
⊤
=
0
3
×
1
(
46
)
\bm{w} \times (c\bm{w}) = c \cdot\bm{w}^\land\bm{w} = -c \cdot(\bm{w}^\top\bm{w}^\land)^\top = \bm{0}_{3\times1} \qquad(46)
w×(cw)=c⋅w∧w=−c⋅(w⊤w∧)⊤=03×1(46)
拉格朗日公式
a
∧
b
∧
=
b
a
⊤
−
(
a
⊤
b
)
I
3
×
3
(
47
)
\bm{a}^\land\bm{b}^\land=\bm{b}\bm{a}^\top-(\bm{a}^\top\bm{b})\mathbf{I}_{3\times3} \qquad (47)
a∧b∧=ba⊤−(a⊤b)I3×3(47)
⇔
a
×
(
b
×
c
)
=
b
(
a
⊤
c
)
−
c
(
a
⊤
b
)
(
48
)
\Leftrightarrow \bm{a} \times (\bm{b} \times \bm{c}) = \bm{b}(\bm{a}^\top\bm{c}) - \bm{c}(\bm{a}^\top\bm{b}) \qquad (48)
⇔a×(b×c)=b(a⊤c)−c(a⊤b)(48)
( a × b ) ∧ = b a ⊤ − a b ⊤ ( 49 ) (\bm{a} \times \bm{b})^\land = \bm{b}\bm{a}^\top-\bm{a}\bm{b}^\top \qquad (49) (a×b)∧=ba⊤−ab⊤(49)
Jacobi Identity
a
×
(
b
×
c
)
+
b
×
(
c
×
a
)
+
c
×
(
a
×
b
)
=
0
(
49
)
\bm{a} \times (\bm{b} \times \bm{c}) + \bm{b} \times (\bm{c} \times \bm{a}) + \bm{c} \times (\bm{a} \times \bm{b}) = \bm{0} \qquad (49)
a×(b×c)+b×(c×a)+c×(a×b)=0(49)
\qquad
or
a
∧
b
∧
c
+
b
∧
c
∧
a
+
c
∧
a
∧
b
=
0
(
50
)
\bm{a}^\land\bm{b}^\land\bm{c} + \bm{b}^\land\bm{c}^\land\bm{a} + \bm{c}^\land\bm{a}^\land\bm{b}= \bm{0} \qquad(50)
a∧b∧c+b∧c∧a+c∧a∧b=0(50)
Rotations
(
R
a
)
∧
=
R
a
∧
R
⊤
(
51
)
p
r
o
o
f
.
(
R
a
)
×
b
=
(
R
a
)
×
(
R
R
⊤
b
)
=
R
[
a
×
(
R
⊤
b
)
]
=
R
a
∧
R
⊤
b
⇒
(
R
a
)
∧
=
R
a
∧
R
⊤
(\bm{R}\bm{a})^\land = \bm{R}\bm{a}^\land\bm{R}^\top \qquad (51) \\ proof.\quad (\bm{Ra}) \times \bm{b} = (\bm{Ra}) \times (\bm{RR^\top b}) = \bm{R}[\bm{a} \times(\bm{R^\top b})]=\bm{R}\bm{a}^\land\bm{R}^\top\bm{b} \\ \Rightarrow (\bm{Ra})^\land = \bm{Ra}^\land\bm{R}^\top
(Ra)∧=Ra∧R⊤(51)proof.(Ra)×b=(Ra)×(RR⊤b)=R[a×(R⊤b)]=Ra∧R⊤b⇒(Ra)∧=Ra∧R⊤
R
(
a
×
b
)
=
(
R
a
)
×
(
R
b
)
(
52
)
\bm{R}(\bm{a} \times \bm{b}) = (\bm{R}\bm{a}) \times (\bm{R}\bm{b}) \qquad(52)
R(a×b)=(Ra)×(Rb)(52)
叉乘反对称矩阵
(
w
∧
)
2
=
w
w
⊤
−
∣
w
∣
2
I
(
53
)
(\bm{w}^\land)^2 = \bm{ww}^\top - |\bm{w}|^2 \mathbf{I} \qquad(53)
(w∧)2=ww⊤−∣w∣2I(53)
( w ∧ ) 3 = ( w w ⊤ − ∣ w ∣ 2 I ) w ∧ = w w ⊤ w ∧ − ∣ w ∣ 2 w ∧ = w ( − w ∧ w ) ⊤ − ∣ w ∣ 2 w ∧ = − ∣ w ∣ 2 w ∧ ( 54 ) \begin{aligned} (\bm{w}^\land)^3 & = (\bm{ww}^\top - |\bm{w}|^2\mathbf{I})\bm{w}^\land\\ & = \bm{w}\bm{w}^\top\bm{w}^\land-|\bm{w}|^2\bm{w}^\land\\ & =\bm{w}(-\bm{w}^\land\bm{w})^\top - |\bm{w}|^2\bm{w}^\land\\ & = -|\bm{w}|^2\bm{w}^\land \end{aligned} \qquad (54) (w∧)3=(ww⊤−∣w∣2I)w∧=ww⊤w∧−∣w∣2w∧=w(−w∧w)⊤−∣w∣2w∧=−∣w∣2w∧(54)
( w ∧ ) 4 = ( w ∧ ) 3 w ∧ = − ∣ w ∣ 2 ( w ∧ ) 2 ( 55 ) \begin{aligned} (\bm{w}^\land)^4 & = (\bm{w}^\land)^3\bm{w}^\land\\ & = -|\bm{w}|^2(\bm{w}^\land)^2 \end{aligned} \qquad (55) (w∧)4=(w∧)3w∧=−∣w∣2(w∧)2(55)
( w ∧ ) 5 = ( w ∧ ) 4 w ∧ = − ∣ w ∣ 2 ( w ∧ ) 3 = − ∣ w ∣ 2 ( − ∣ w ∣ 2 w ∧ ) = ∣ w ∣ 4 w ∧ ( 56 ) \begin{aligned} (\bm{w}^\land)^5 & = (\bm{w}^\land)^4\bm{w}^\land \\ &=-|\bm{w}|^2(\bm{w}^\land)^3\\ &=-|\bm{w}|^2(-|\bm{w}|^2\bm{w}^\land)\\ &=|\bm{w}|^4\bm{w}^\land \qquad (56) \end{aligned} (w∧)5=(w∧)4w∧=−∣w∣2(w∧)3=−∣w∣2(−∣w∣2w∧)=∣w∣4w∧(56)
( w ∧ ) 6 = ( w ∧ ) 5 w ∧ = ∣ w ∣ 4 ( w ∧ ) 2 ( 57 ) \begin{aligned} (\bm{w}^\land)^6 & = (\bm{w}^\land)^5\bm{w}^\land\\ &= |\bm{w}|^4(\bm{w}^\land)^2 \qquad(57) \end{aligned} (w∧)6=(w∧)5w∧=∣w∣4(w∧)2(57)
( w ∧ ) 7 = ( w ∧ ) 6 w ∧ = ∣ w ∣ 4 ( w ∧ ) 3 = ∣ w ∣ 4 ( − ∣ w ∣ 2 w ∧ ) = − ∣ w ∣ 6 w ∧ ( 58 ) . . . . . . \begin{aligned} (\bm{w}^\land)^7 &= (\bm{w}^\land)^6\bm{w}^\land\\ &=|\bm{w}|^4(\bm{w}^\land)^3\\ &=|\bm{w}|^4(-|\bm{w}|^2\bm{w}^\land)\\ &=-|\bm{w}|^6\bm{w}^\land \qquad(58)\\ &...... \end{aligned} (w∧)7=(w∧)6w∧=∣w∣4(w∧)3=∣w∣4(−∣w∣2w∧)=−∣w∣6w∧(58)......
》1.3.2 Properties of the matrix
Ω
\mathbf{\Omega}
Ω
Ω
\mathbf{\Omega}
Ω矩阵出现在一个向量与四元素的乘积当中,可用于四元素的求导,它具有以下性质:
Ω
(
w
)
=
[
0
−
w
x
−
w
y
−
w
z
w
x
0
w
z
−
w
y
w
y
−
w
z
0
w
x
w
z
w
y
−
w
x
0
]
=
[
0
−
w
⊤
w
−
w
∧
]
(
59
)
\begin{aligned} \mathbf{\Omega}(\bm{w}) &=\begin{bmatrix} 0 & -w_x & -w_y & -w_z\\ w_x & 0 & w_z & -w_y\\ w_y & -w_z& 0 & w_x \\ w_z & w_y & -w_x & 0 \end{bmatrix}\\ & = \begin{bmatrix} 0 & -\bm{w}^\top\\ \bm{w} & -\bm{w}^\land \end{bmatrix} \end{aligned} \qquad (59)
Ω(w)=⎣⎢⎢⎡0wxwywz−wx0−wzwy−wywz0−wx−wz−wywx0⎦⎥⎥⎤=[0w−w⊤−w∧](59)
Ω ( w ) 2 = [ − w ⊤ w w ⊤ w ∧ − w ∧ w − w w ⊤ + w ∧ w ∧ ] = [ − ∣ ∣ w ∣ ∣ 2 0 1 × 3 0 3 × 1 − ∣ ∣ w ∣ ∣ 2 I 3 × 3 ] = − ∣ ∣ w ∣ ∣ 2 I 4 × 4 ( 60 ) \begin{aligned} \mathbf{\Omega}(\bm{w})^2 & = \begin{bmatrix} -\bm{w}^\top\bm{w} & \bm{w}^\top\bm{w}^\land\\ -\bm{w}^\land\bm{w} & -\bm{ww}^\top+\bm{w}^\land\bm{w}^\land\\ \end{bmatrix}\\ &=\begin{bmatrix} -||\bm{w}||^2 & \mathbf{0}_{1\times3} \\ \mathbf{0}_{3\times1}& -||\bm{w}||^2\mathbf{I}_{3\times3} \end{bmatrix}\\ &=-||\bm{w}||^2\mathbf{I}_{4\times4} \qquad\qquad\qquad\qquad (60) \end{aligned} Ω(w)2=[−w⊤w−w∧ww⊤w∧−ww⊤+w∧w∧]=[−∣∣w∣∣203×101×3−∣∣w∣∣2I3×3]=−∣∣w∣∣2I4×4(60)
Ω ( w ) 3 = − ∣ ∣ w ∣ ∣ 2 Ω ( w ) ( 61 ) \mathbf{\Omega}(\bm{w})^3 = -||\bm{w}||^2\mathbf{\Omega}(\bm{w}) \qquad(61) Ω(w)3=−∣∣w∣∣2Ω(w)(61)
Ω ( w ) 4 = ∣ ∣ w ∣ ∣ 4 I 4 × 4 ( 62 ) \mathbf{\Omega}(\bm{w})^4 = ||\bm{w}||^4\mathbf{I}_{4\times4} \qquad(62) Ω(w)4=∣∣w∣∣4I4×4(62)
Ω ( w ) 5 = ∣ ∣ w ∣ ∣ 4 Ω ( w ) ( 63 ) \mathbf{\Omega}(\bm{w})^5 = ||\bm{w}||^4\mathbf{\Omega}(\bm{w}) \qquad(63) Ω(w)5=∣∣w∣∣4Ω(w)(63)
Ω ( w ) 6 = − ∣ ∣ w ∣ ∣ 6 I 4 × 4 ( 64 ) \mathbf{\Omega}(\bm{w})^6 = -||\bm{w}||^6\mathbf{I}_{4\times4} \qquad(64) Ω(w)6=−∣∣w∣∣6I4×4(64)
1.3.3 Properties of the matrix Ξ \mathbf{\Xi} Ξ
Ψ ( q ) = [ − q v ⊤ q w I 3 × 3 + q v ∧ ] , Ψ ⊤ ( q ) = [ − q v q w I 3 × 3 − q v ∧ ] ( 65 ) \mathbf{\Psi}(\bm{q}) = \begin{bmatrix} -\bm{q}_v^\top \\ q_w\mathbf{I}_{3\times3} + \bm{q}_v^\land \end{bmatrix}, \quad\mathbf{\Psi}^\top(\bm{q})=[-\bm{q}_v\quad q_w\mathbf{I}_{3\times3}-\bm{q}_v^\land] \qquad (65) Ψ(q)=[−qv⊤qwI3×3+qv∧],Ψ⊤(q)=[−qvqwI3×3−qv∧](65)
⇒ Ψ ⊤ ( q ) Ψ ( q ) = ∣ ∣ q ∣ ∣ 2 I 3 × 3 ( 66 ) Ψ ( q ) Ψ ⊤ ( q ) = ∣ ∣ q ∣ ∣ I 4 × 4 − q q ⊤ ( 67 ) Ψ ⊤ ( q ) q = 0 3 × 1 ( 68 ) \begin{aligned} \Rightarrow \qquad \mathbf{\Psi}^\top(\bm{q})\mathbf{\Psi}(\bm{q}) & = ||\bm{q}||^2\mathbf{I}_{3\times3} \qquad(66)\\ \mathbf{\Psi}(\bm{q})\mathbf{\Psi}^\top(\bm{q}) &= ||\bm{q}||\mathbf{I}_{4\times4}-\bm{q}\bm{q}^\top\qquad(67)\\ \mathbf{\Psi}^\top(\bm{q})\bm{q}&=\mathbf{0}_{3\times1} \qquad (68) \end{aligned} ⇒Ψ⊤(q)Ψ(q)Ψ(q)Ψ⊤(q)Ψ⊤(q)q=∣∣q∣∣2I3×3(66)=∣∣q∣∣I4×4−qq⊤(67)=03×1(68)
Ξ
\mathbf{\Xi}
Ξ与
Ω
\mathbf{\Omega}
Ω之间的关系为:
Ω
(
w
)
q
=
Ψ
(
q
)
w
(
69
)
\mathbf{\Omega}(\bm{w})\bm{q}=\mathbf{\Psi}(\bm{q})\bm{w} \qquad(69)
Ω(w)q=Ψ(q)w(69)
》1.4 四元素与旋转矩阵之间的关系
给定旋转四元素
q
=
cos
(
θ
/
2
)
+
u
sin
(
θ
/
2
)
\bm{q}=\cos(\theta/2)+\bm{u}\sin(\theta/2)
q=cos(θ/2)+usin(θ/2),对应的旋转矩阵记为
R
(
q
)
\mathbf{R}(\bm{q})
R(q),则
R
(
q
)
=
(
2
q
w
2
−
1
)
I
+
2
q
v
q
v
⊤
+
2
q
w
q
v
∧
(
70
)
\mathbf{R}(\bm{q})=(2q_w^2-1)\mathbf{I} + 2\bm{q}_v\bm{q}_v^\top + 2q_w\bm{q}_v^\land \qquad(70)
R(q)=(2qw2−1)I+2qvqv⊤+2qwqv∧(70)
其中:
q
w
=
cos
(
θ
/
2
)
,
q
v
=
u
sin
(
θ
/
2
)
q_w = \cos(\theta/2), \quad\bm{q}_v=\bm{u}\sin(\theta/2)
qw=cos(θ/2),qv=usin(θ/2)
或者表示为:
R
(
q
)
=
Ξ
⊤
(
q
)
Ψ
(
q
)
(
71
)
\mathbf{R}(\bm{q})=\mathbf{\Xi}^\top(\bm{q})\mathbf{\Psi}(\bm{q}) \qquad(71)
R(q)=Ξ⊤(q)Ψ(q)(71)
或者表示为:
R
(
q
)
=
[
1
−
2
q
y
2
−
2
q
z
2
2
q
x
q
y
−
2
q
w
q
z
2
q
x
q
z
+
2
q
w
q
y
2
q
x
q
y
+
2
q
w
q
z
1
−
2
q
x
2
−
2
q
z
2
2
q
y
q
z
−
2
q
w
q
x
2
q
x
q
z
−
2
q
w
q
y
2
q
y
q
z
+
2
q
w
q
x
1
−
2
q
x
2
−
2
q
y
2
]
(
72
)
\mathbf{R}(\bm{q}) = \begin{bmatrix} 1-2q_y^2-2q_z^2 & 2q_xq_y - 2q_wq_z & 2q_xq_z + 2q_wq_y \\ 2q_xq_y+2q_wq_z & 1-2q_x^2-2q_z^2 & 2q_yq_z-2q_wq_x\\ 2q_xq_z-2q_wq_y& 2q_yq_z+2q_wq_x& 1-2q_x^2-2q_y^2 \end{bmatrix} \qquad(72)
R(q)=⎣⎡1−2qy2−2qz22qxqy+2qwqz2qxqz−2qwqy2qxqy−2qwqz1−2qx2−2qz22qyqz+2qwqx2qxqz+2qwqy2qyqz−2qwqx1−2qx2−2qy2⎦⎤(72)
旋转矩阵的乘法与旋转四元素的乘法之间的关系:
R
(
0
q
1
)
R
(
1
q
2
)
=
R
(
0
q
1
⊗
1
q
2
)
(
73
)
\mathbf{R}(\bm{^0q_1})\mathbf{R}(^1\bm{q}_2) = \mathbf{R}(^0\bm{q}_1 \otimes {^1}\bm{q}_2) \qquad(73)
R(0q1)R(1q2)=R(0q1⊗1q2)(73)
指数映射
R
(
q
)
=
e
x
p
(
θ
u
∧
)
(
74
)
\mathbf{R}(\bm{q}) = exp(\theta\bm{u}^\land) \qquad (74)
R(q)=exp(θu∧)(74)
》1.5 四元素对时间的导数
待续。。。。。。
参考文献:
[1] Indirect Kalman Filter for 3D Attitude Estimation
[2] Quaternion kinematics for the error-state Kalman filter
[3] 视觉SLAM十四讲 从理论到实践