Meccanica (2024) 59:89–106

https://doi.org/10.1007/s11012-023-01741-7

Analyzing the angular acceleration vector of a moving rigid

body
J. Jesús Cervantes‑Sánchez · José M. Rico‑Martínez ·
Mario A. García‑Murillo · Marco A. Gómez‑Álvarez

Received: 28 June 2023 / Accepted: 25 November 2023 / Published online: 31 December 2023
© Springer Nature B.V. 2023

Abstract This paper presents a novel and system- techniques or advanced mathematical tools. Due to
atic approach for obtaining the angular acceleration its vector nature, the expression for the angular accel-
vector of a moving rigid body. The novelty of the eration vector proposed in this paper is relatively sim-
proposed method lies in the particular form of writ- ple, as well as, it is very robust against computational
ing the pose of the moving rigid body, as well as in singularities. Two fully detailed case studies demon-
the procedure to compute its time derivatives. The strate the robustness of the proposed angular acceler-
derivation process goes directly to the very founda- ation vector compared with other expressions appear-
tions of rotational motion and exploits the phenom- ing in the literature.
enological connection between orientation, angular
velocity, angular acceleration, and spatial motion of Keywords Angular acceleration vector · Spatial
a rigid body. Hence, as a remarkable result, a sym- motion · Rigid body · Non-collinear points
bolic expression for the angular acceleration vector
arises naturally without the need to solve the inverse
acceleration problem. The novel and general expres- 1 Introduction
sion of the angular acceleration vector involves rela-
tionships between the position, velocity, and accelera- There is a phenomenological connection between ori-
tion vectors of three non-collinear points of the body, entation, angular velocity, and angular acceleration,
which can be easily understood and physically inter- which are strongly associated with the spatial motion
preted without particular knowledge of specialized of a rigid body. The formal study of rigid body rota-
tions may be traced to 1775 when Euler published
J. J. Cervantes‑Sánchez (*) · J. M. Rico‑Martínez · his seminal work [1], which was rediscovered inde-
M. A. García‑Murillo · M. A. Gómez‑Álvarez pendently by Rodrigues [2] in 1840. Since then, any
Departamento de Ingeniería Mecánica, Universidad number of discoveries [3–6] have flowed continu-
de Guanajuato, DICIS, 36885 Salamanca, Guanajuato,
ously from one author to the next, all of them add-
Mexico
e-mail: jecer@ugto.mx ing something new to the results obtained by their
predecessors.
J. M. Rico‑Martínez
e-mail: jrico@ugto.mx A systematic formulation of the dynamic model of
a mechanical system is needed to predict and under-
M. A. García‑Murillo
e-mail: garcia.mario@ugto.mx stand its behavior. The effectiveness in formulating
the equations of motion depends primarily on the
M. A. Gómez‑Álvarez
e-mail: marco.gomez@ugto.mx ability to construct simple and correct mathematical

Vol.: (0123456789)
13
 90 Meccanica (2024) 59:89–106

expressions for kinematic quantities such as angular kinematic property related to the acceleration state of
velocities and accelerations of rigid bodies, as well a moving rigid body, which arises naturally when a
as, velocities and accelerations of points of moving proper description of the spatial motion of a rigid body
rigid bodies [7]. For example, Euler’s second law is carried out.
requires a proper formulation of the angular velocity
vector and angular acceleration vector of each rigid
body composing a spatial mechanical system moving 2 Description of the spatial motion of a moving
in a three-dimensional space [8]. One difficulty with rigid body
the study of the angular velocity vector, and the angu-
lar acceleration vector, is that it becomes increasingly To start with, it is reasonable to think that to obtain a
difficult to formulate as the complexity of the motion good mathematical model of the angular acceleration,
of the rigid body increases, such as occurs with the one must first make a careful description and analysis
links of spatial parallel manipulators [9], flight simu- of both, the orientation and the angular velocity of the
lators [10], and complex machines, e.g., the turbula body. On the other hand, a moving rigid body may be
machine [11], or the human head [12]. translating and rotating simultaneously in a general spa-
A fundamental problem in rigid body kinematics is tial motion. Hence, it is important to correlate the angu-
the inverse acceleration problem [13]. This is a very lar motion of the body with the translational motion of
challenging problem for spatial motion that consists any point of the moving body. In this regard, we take
of the determination of the angular acceleration vec- some ideas from previous investigations [22], which
tor in terms of the position, velocity, and accelera- deal with the angular velocity of a rigid body in motion.
tion vectors of three non-collinear points of a moving
rigid body. As far as we know, very few attempts have 2.1 The pose of a moving rigid body
been reported to solve this problem. Among these
investigations, regarding vector-based approaches, it Consider the rigid body shown in Fig. 1, which may be
is fair to highlight the contributions of Condurache moving in any manner with respect to a fixed reference
and Matcovschi [13], Field and Ziwet [14], Soutas- frame XYZ. The body has three arbitrary and non-col-
Little and Inman [15], and Wittenburg [16, 17]. On linear points, namely, points 1, 2, and 3, whose location
the other hand, Angeles [18–20], and Condurache and with respect to the origin O of the fixed frame XYZ is
Matcovschi [21] address the same problem, but with given by position vectors, p1, p2, and p3, respectively.
a matrix-based approach. The pose of the rigid body shown in Fig. 1 can be
All the derivations [13–21] for the angular accel- described in terms of the location and orientation of the
eration vector have much in common and, to a greater moving frame UVW, fixed to the body, with respect to
or lesser extent, all of them follow a general pattern: the fixed frame XYZ. On one hand, the location of the
(a) Start from classical and well-known equations origin of the moving frame UVW may be defined by
related to the acceleration state of a rigid body, (b) It the position vector of point 1, namely, p1. On the other
is required to solve the inverse acceleration problem, hand, the orientation of the body is completely deter-
(c) They do not provide further details about the intrin- mined once the set of coordinate axes UVW has been
sic nature of the angular acceleration, (d) Involve the oriented relative to the fixed reference frame XYZ. Both
angular velocity, which, in turn, comes from the veloc- requirements may be stated in terms of the position
ity state of a moving rigid body, (e) The denominators vectors p1, p2, and p3, of the three non-collinear points
of the resulting expressions are prone to computa- 1, 2, and 3, respectively, which are shown in Fig. 1. To
tional singularities, and (f) Suffer from a rather heavy this end, we define the following unit vectors:
computational burden associated with all the required p 2 − p1 p − p1
matrix computations, which significantly obscures the u≡ √ = 2 (1)
(p2 − p1 ) ⋅ (p2 − p1 ) L12
geometrical nature of the angular acceleration vector.
Hence the motivation to devise an alternative and more
comprehensive approach that overcomes the shortcom- p 3 − p1 p − p1
m≡ √ = 3 (2)
ings (a)-(f) mentioned above. Furthermore, the objec- (p3 − p1 ) ⋅ (p3 − p1 ) L13
tive is to exploit the fact that angular acceleration is a

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 91

Fig. 1  Fixed (XYZ) and

moving (UVW) frames used
for describing the pose of a
rigid body

m − (m ⋅ u)u 1 m⋅u However, its Cartesian coordinates (measured in the

v≡ = k1 m − k2 u, k1 ≡ , k2 ≡ .
𝜆 𝜆 𝜆 fixed frame XYZ), namely, x, y, and z, are continu-
(3) ously changing, whereas rotational motion occurs. In
this way, the position vector r can be expressed with
w≡u×v respect to the moving frame as follows:
(4) r = uu + vv + ww
where the scalar parameter:
(7)
√
𝜆 ≡ {m − (m ⋅ u)u} ⋅ {m − (m ⋅ u)u} (5)
Finally, substitution of Eq. (7) into Eq. (6) yields the
is graphically depicted in Fig. 1. following result:
It is important to remark that unit vectors u, v, and
w are directed along the axes U, V, and W, respec-
rP∕O = p1 + u u + v v + w w. (8)
tively, and they are used to describe the relative orien- which is a vector equation1 related to an arbitrary
tation between frames UVW and XYZ. Furthermore, pose of the moving rigid body depicted in Figs. 1, and
unit vectors u, and m are directly related to the the 2. This equation clearly shows that position vector p1,
position vectors, p1, p2, and p3, whereas unit vectors and unit vectors u, v, and w can be used to locate all
v, and w can be computed in terms of unit vectors u, the points of the body in space. Moreover, the Eq. (8)
and m. is the key equation to correlate the angular motion of
We may interpret the pose of a body as a way to the body, which is represented by rotating unit vectors
know about the position of all the points of the body u, v, and w, with the translational motion of any point
in the space. In this regard, consider a typical point of the moving body. Furthermore, as will be seen
P, fixed in the body, which is shown in Figs. 1 and later, this equation will lead to a clear, simple, and
2. Thus, one may state that the position of point P, systematic way to obtain the velocity and acceleration
as seen from XYZ, equals the position of point 1, as state of the rigid body in motion.
seen from XYZ, plus the position of point P relative to
point 1, that is: 2.2 The first time derivative of body pose
rP∕O = p1 + r (6)
Since the body-fixed UVW frame translates and
where position vector r is used to represent the posi- rotates relative to the XYZ frame, the position vec-
tion of an arbitrary point P of the moving body with tor p1, as well as unit vectors u, v and w will change
respect to the origin of the moving frame UVW. Thus,
1
position vector r maintains a constant magnitude and It should be noted that Eq. (8) can be equivalently written
as rP∕O = p1 + R r , where the (3 × 3) matrix R ≡ [u v w]
orientation in the UVW frame, that is, its coordinates describes the orientation of the rigid body with respect to ref-
u, v, and w remain fixed even if the body rotates. erence frame XYZ, and r ≡ (u, v, w)T .

Vol.: (0123456789)
13
 92 Meccanica (2024) 59:89–106

derivative, we will use a dot over a parameter to

indicate a time derivative of that parameter, e.g.,
d a∕d t ≡ ȧ .
Next, by projecting each time derivative of the
involved unit vectors onto the axes of the UVW frame,
we may write:
u̇ = (u̇ ⋅ u)u + (u̇ ⋅ v)v + (u̇ ⋅ w)w
v̇ = (v̇ ⋅ u) u + (v̇ ⋅ v) v + (v̇ ⋅ w) w (12)
ẇ = (ẇ ⋅ u)u + (ẇ ⋅ v)v + (ẇ ⋅ w)w

Furthermore, unit vectors must satisfy the following

relationships:
u ⋅ u = 1, v ⋅ v = 1, w ⋅ w = 1. (13)
Fig. 2  Orientation of moving frame UVW with respect to fixed as well as:
frame XYZ
u ⋅ v = 0, u ⋅ w = 0, v ⋅ w = 0. (14)
through time, and they may be considered as func- Then, time differentiation of Eqs. (13) and (14) leads
tions of time, denoted by the symbol t. Therefore, this to:
section starts by taking the first time derivative of the
u̇ ⋅ u = 0, v̇ ⋅ v = 0, ẇ ⋅ w = 0, (15)
vector Eq. (6), which is given by:
( ) ( ) ( )
d rP∕O d p1 dr u̇ ⋅ v = −u ⋅ v,
̇ u̇ ⋅ w = −u ⋅ w, ̇ (16)
̇ v̇ ⋅ w = −v ⋅ w.
= + (9)
dt d t d t XYZ
XYZ XYZ Thus, Eq. (12) become:
where (d∕dt)XYZ is used to denote the time derivative u̇ = −(u ⋅ v)v
̇ + (u̇ ⋅ w)w
as seen from XYZ.
v̇ = +(v̇ ⋅ u) u − (v ⋅ w)
̇ w (17)
On the one hand, the first two terms of Eq. (9) have
a direct physical interpretation. The time rate change ẇ = −(u̇ ⋅ w)u + (v ⋅ w)v
̇
of rP∕O, as seen from XYZ, represents the veloc- Substituting Eq. (17) into Eq. (11) we find that:
ity vector of point P with respect to fixed point O,
namely, vP∕O, whereas the time rate change of p1 with ṙ = {v(v⋅u)−w(
̇ u⋅w)}u+{w(v⋅
̇ w)−u(u⋅
̇ v)}v
̇
(18)
respect to fixed point O, as seen from XYZ, represents +{u(u⋅w)−v(v⋅
̇ w)}w
̇
the velocity vector of point 1, namely, v1∕O. Thus, we
can express the foregoing statements as follows: By resorting to the definition of the cross product
between two vectors, one may notice that Eq. (18) can
( ) ( )
d rP∕O d p1 be written as follows:
vP∕O ≡ , v1∕O ≡ . (10)
dt XYZ
d t XYZ ṙ = −{(v⋅ w)u
̇ + (u⋅w)v
̇ + (u⋅ v)w}
̇ × {u u + v v + w w}
(19)
On the other hand, the second term appearing on the
left-hand side of the Eq. (9) deserves special treat- Thus, from Eqs. (7) and (16), Eq. (19) becomes:
ment. To this end, and recalling Eq. (7), we have that: ṙ = {(v⋅w)u
̇ − (u⋅w)v
̇ + (u⋅v)w}
̇ ×r (20)
( )
dr where first time derivatives of unit vectors, u̇ , v̇ and
= u u̇ + v v̇ + w ẇ (11)
d t XYZ ẇ , can be computed by taking the first time derivative
of Eq. (1)–(4), respectively, thus yielding:
since the components u, v and w do not change
through the time. Moreover, when it is clear from
the discussion what frame is involved for a time

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 93

ṗ 2 − ṗ 1
u̇ = (21) n ≡ (v⋅w)u
̇ − (u⋅w)v
̇ + (u⋅v)w.
̇ (29)
L12
ṅ = (v⋅w)u
̈ + (v⋅ ̇ + (v⋅w)
̇ w)u ̇ u̇ − (u⋅w)v
̈ − (u⋅ ̇ − (u⋅w)
̇ w)v ̇ v̇
ṗ − ṗ 1
ṁ = 3 (22)
+ (u⋅v)w
̈ + (u⋅ ̇ + (u⋅v)
̇ v)w ̇ w. ̇
L13 ṅ ≡ ṅ 1 + ṅ 2 + ṅ 3 − ṅ 4 − ṅ 5 − ṅ 6 + ṅ 7 + ṅ 8 + ṅ 9 .
(30)
v̇ =k1 ṁ − k2 u̇ (23) and the second time derivative of unit vectors,
namely, ü , v̈ , and w
̈ , can be computed by taking the
ẇ =u̇ × v + u × v̇ (24) first time derivative of Eqs. (21)–(24), respectively,
thus yielding:
where it is important to remark that unit vectors u̇ ,
and ṁ are directly related to the velocity vectors of p̈ 2 − p̈ 1
points 1, 2, and 3, namely, ṗ 1, ṗ 2, and ṗ 3, whereas ü = (31)
L12
vectors v̇ , and ẇ can be computed in terms of vectors
u̇ , and ṁ .
p̈ 3 − p̈ 1
Finally, from Eqs. (10), and (20), Eq. (9) becomes: ̈ =
m (32)
L13
vP∕O = v1∕O + {(v⋅w)
̇ u − (u⋅w)
̇ v + (u⋅v)
̇ w} × r.
(25) ̈ − k2 ü
v̈ = k1 m (33)
which may be considered as the vector equation rep-
resenting the first-time derivative of the body pose ̈ = ü × v + 2 u̇ × v̇ + u × v.
w ̈ (34)
illustrated in Fig. 1. This equation is closely related
to the velocity state [23] of the rigid body since it pro- where it is important to remark that unit vectors ü ,
vides enough information to find the velocity of any and m ̈ are directly related to the acceleration vectors
point of the moving rigid body. of points 1, 2, and 3, namely, p̈ 1, p̈ 2, and p̈ 3, whereas
vectors v̈ , and w ̈ can be computed in terms of vectors
ü , and m ̈.
2.3 The second time derivative of body pose Equation (30) involves unit vectors u, v, w, as
well as, their first, and second time derivatives,
Both sides of the vector Eq. (25) may be differenti- and it is composed of nine vector terms denoted as
ated with respect to time to obtain the acceleration ṅ 1 , ṅ 2 , ⋯ ṅ 9. In order to improve the readability of the
equation, which is given by: article, a detailed computation of each term is pre-
sented in Appendix 1. As shown in that appendix, the
aP∕O = a1∕O + aP∕1 (26) general idea is to include only u, u̇ , ü , m, ṁ , and m ̈,
since these vectors are directly related to the position,
where
the velocity, and the acceleration of the three non-col-
( ) ( ) linear points of the moving rigid body under analysis.
d vP∕O d v1∕O
aP∕O ≡ , a1∕O ≡ . (27)
dt XYZ
dt XYZ

represents the acceleration vector of point P with 3 The angular acceleration vector
respect to fixed point O, and the acceleration vector
of point 1 with respect to fixed point O, respectively, We are now in a position to derive the expression for
whereas the second term of the left-hand side of Eq. the angular acceleration vector. To this end, the gen-
(26) deserves special treatment, that is: eral Eq. (26) is combined with Eqs. (20), (28), and
( ) (29), thus leading to:
d ṙ
aP∕1 ≡ = ṅ × r + n × r.
̇ (28) aP∕O = a1∕O + ṅ × r + n × (n × r). (35)
d t XYZ

where, we have that:

Vol.: (0123456789)
13
 94 Meccanica (2024) 59:89–106

which is an equation closely related to the accelera- (b) It involves only the position, velocity, and accel-
tion state [23] of the moving rigid body since it pro- eration of three non-collinear points of a moving
vides enough information to find the acceleration of rigid body.
any point of the moving rigid body. (c) It does not require the computation of the angu-
A dimensional analysis reveals that Eq. (35) has lar velocity vector of the moving body. All the
the dimensions of acceleration, and vector ṅ has the expressions for the angular acceleration reported
same units as the angular acceleration. Moreover, in [14, 15, 17, 20], require a previous computa-
vector ṅ is indeed the angular acceleration vector, tion of the corresponding angular velocity vector
whereas vector n is the angular velocity vector. Fur- of the moving body.
thermore, the angular acceleration vector is usually (d) Its denominator is given by a simple scalar
denoted by the bold Greek symbol 𝜶 . Hence, from parameter, namely, 𝜆2, see Eqs. (3), and (5).
this point, ṅ ≡ 𝜶 . Thus, a careful collection of all Moreover, it can be proved that 𝜆 = sin 𝛾 , where
the nine terms shown in Eq. (30) yields the follow- 𝛾 is the angle formed by unit vectors u, and m,
ing result: see Fig. 1. Hence, the parameter 𝜆 only vanishes
when the three given points, 1, 2, and 3, are col-
𝜶 = k12 {𝜿 1 + 𝜿 2 + 𝜿 3 + 𝜿 4 + 𝜿 5 + 𝜿 6 + 𝜿 7 } − k1 k2 {𝜿 8 + 𝜿 9 }.
linear, i.e., 𝛾 = 0◦, or 𝛾 = 180◦, which is not the
(36)
general case treated in this paper. Therefore, it is
where we have the following scalar parameters: very robust against computational singularities.
� �
1
�
m⋅u
� (e) It has a simple mathematical structure since only
k1 ≡ , k2 ≡ , basic principles of vector calculus were used in
𝜆 𝜆
(37) its formulation.
√
𝜆≡ {m − (m ⋅ u)u} ⋅ {m − (m ⋅ u)u}.

as well as the following vectors: 4 Representations of the angular acceleration

𝜿 1 ≡ (m×u)×(u×m)
̈ vector
𝜿 2 ≡ ü × {(u × m) × m}
In reviewing the literature we found five typical repre-
𝜿 3 ≡ {ṁ ⋅ (u̇ × m)}u sentations for the angular acceleration vector, which are
𝜿 4 ≡ {ṁ ⋅ (u × m)}u̇ different from the one given by the Eq. (36). All those
{ }
𝜿 5 ≡ u×
̇ (u×m)× ṁ (38) representations are required to solve the inverse accel-
{ } eration problem in rigid body kinematics [13], and they
𝜿 6 ≡ u×
̇ (u×m)×m
̇
{ } are briefly presented next for completeness purposes.
𝜿 7 ≡ u×
̇ (u× m)×m
̇
𝜿 8 ≡ (u⋅u)(u×m)
̈ 4.1 First representation of the angular acceleration
𝜿 9 ≡ (u⋅
̇ u)(u×m)
̇ vector

whose detailed derivations are presented in Appendix A first representation of the angular acceleration vector
1. is due to Field, and Ziwet [14], which is given by:
It is important to remark that Eq. (36) is a novel ( ) ( )
and general expression for computing the angular 𝜶 FW =
𝝎FW ⋅ q̈
p−
𝝎FW ⋅ p̈
q
acceleration vector of a rigid body moving in space. 𝝎FW ⋅ (p × q) 𝝎FW ⋅ (p × q)
( )
As far as we know, this equation and its detailed (𝝎FW ⋅ p)(𝝎FW ⋅ q)
+ 𝝎FW −
derivation have not been reported previously in the 𝝎FW ⋅ (p × q)
literature. This equation has remarkable features, ( ) ( )
(𝝎FW ⋅ 𝝎FW )(p ⋅ q) p ⋅ q̈
such as: −
𝝎FW ⋅(p×q)
𝝎FW −
𝝎FW ⋅(p×q)
𝝎FW .

𝜶 FW ≡ s1 − s2 + s3 − s4 − s5 .
(a) This is a result that arose naturally, without the
(39)
need to solve the inverse acceleration problem.

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 95

where q2∕3 ≡ p̈ 2 − p̈ 3 − 𝝎W × (ṗ 2 − ṗ 3 ). (49)

ṗ × q̇
𝝎FW = (40) r3∕1 ≡ p3 − p1 .
ṗ ⋅ q (50)

is the angular velocity vector of the moving rigid However, when the denominator of Eq. (47) equals
body, and p ≡ p2 − p1, q ≡ p3 − p1, ṗ ≡ ṗ 2 − ṗ 1, zero, in [17] it is proposed the following alternative
q̇ ≡ ṗ 3 − ṗ 1, p̈ ≡ ṗ 2 − p̈ 1, and q̈ ≡ p̈ 3 − p̈ 1. formula:
𝜶 W2 = 𝜇1 (p1 − p2 ) + 𝜇2 (p2 − p3 ), for r3∕1 ⋅ q2∕3 = 0.
4.2 Second representation of the angular acceleration
(51)
vector
where
A second representation for the angular acceleration ( )
1
vector has been proposed by Soutas-Little and Inman 𝜇1 ≡ { 𝜺 ⋅ [ p̈ 3 − p̈ 2 − 𝝎W × (ṗ 3 − ṗ 2 ) ] }.
𝜺⋅𝜺
[15]. The corresponding formula is as follows:
(52)
qB∕A × qC∕A ( )
1
𝜶 SLI =
qB∕A ⋅ rC∕A
, for qB∕A ⋅ rC∕A ≠ 0. (41) 𝜇2 ≡ { 𝜺 ⋅ [ p̈ 1 − p̈ 2 − 𝝎W × (ṗ 1 − ṗ 2 ) ] }.
𝜺⋅𝜺
(53)
where
𝜺 ≡ (p1 − p2 ) × (p3 − p2 ). (54)
qB∕A ≡ p̈ 2 − p̈ 1 − 𝝎SLI × {𝝎SLI × (p2 − p1 )}. (42)
The corresponding angular velocity vector, namely,
𝝎W , is given by:
qC∕A ≡ p̈ 3 − p̈ 1 − 𝝎SLI × {𝝎SLI × (p3 − p1 )}. (43)
(ṗ 1 − ṗ 3 ) × (ṗ 2 − ṗ 3 )
𝝎W1 = ,
rC∕A ≡ p3 − p1 . (44) (p3 − p1 ) ⋅ (ṗ 2 − ṗ 3 ) (55)
for (p3 − p1 ) ⋅ (ṗ 2 − ṗ 3 ) ≠ 0,
and 𝝎SLI denotes the angular velocity vector, which is
given by:
𝝎W2 = 𝜇3 (p1 − p2 ) + 𝜇4 (p2 − p3 ),
(56)
for (p3 − p1 ) ⋅ (ṗ 2 − ṗ 3 ) = 0.
vB∕A × vC∕A
𝝎SLI =
vB∕A ⋅ rC∕A
, for vB∕A ⋅ rC∕A ≠ 0. (45)
where
being: ( )
𝜇3 ≡
1
vB∕A ≡ ṗ 2 − ṗ 1 , vC∕A ≡ ṗ 3 − ṗ 1 . (46) { 𝜺 ⋅ (ṗ 3 − ṗ 2 ) },
𝜺⋅𝜺
( ) (57)
𝜇4 ≡
1
4.3 Third representation of the angular acceleration { 𝜺 ⋅ (ṗ 1 − ṗ 2 ) }.
𝜺⋅𝜺
vector

Professor Wittenburg proposed a third representation of 4.4 Fourth representation of the angular acceleration
the angular acceleration vector [17]. The corresponding vector
formula is as follows:
A fourth representation of the angular acceleration vec-
q1∕3 × q2∕3
𝜶 W1 = , for r3∕1 ⋅ q2∕3 ≠ 0. (47) tor was developed by Angeles [20], and the correspond-
r3∕1 ⋅ q2∕3 ing formulas are given by:
where 𝜶 A1 = D−1 vec (P̈ − 𝛀21 P), for tr(P) ≠ 0. (58)
q1∕3 ≡ p̈ 1 − p̈ 3 − 𝝎W × (ṗ 1 − ṗ 3 ). (48) where

Vol.: (0123456789)
13
 96 Meccanica (2024) 59:89–106

2 4
D−1 = 𝛼1 I − 𝛽 P2 , 𝛼1 ≡ , 𝛽≡ 2
,
tr(P) tr(P)(tr(P ) − tr2 (P))
� �
P ≡ p1 − c p2 − c p3 − c ,
� � � �
Ṗ ≡ [ṗ ij ] = ṗ 1 − ċ ṗ 2 − ċ ṗ 3 − ċ , P̈ ≡ p̈ 1 − c̈ p̈ 2 − c̈ p̈ 3 − c̈ ,
p + p2 + p3 ṗ + ṗ 2 + ṗ 3 p̈ + p̈ 2 + p̈ 3
c≡ 1 , ċ ≡ 1 , c̈ ≡ 1 ,
3 3 3
⎡ ṗ 32 − ṗ 23 ⎤ ⎡ 𝜔1 ⎤
̇ ≡ 1 ⎢ ṗ 13 − ṗ 31 ⎥, 𝝎A1 ≡ ⎢ 𝜔2 ⎥ = D−1 vec(P),
vec(P) ̇
2⎢ ⎥ ⎢ ⎥
⎣ ṗ 21 − ṗ 12 ⎦ ⎣ 𝜔3 ⎦
⎡ 0 − 𝜔3 𝜔2 ⎤
𝛀 1 ≡ ⎢ 𝜔3 0 − 𝜔1 ⎥.
⎢ ⎥
⎣ −𝜔2 𝜔 1 0 ⎦

and tr(P) denotes the trace of a (3 × 3) matrix P , and I p̈ Q = [ p̈ 1 ⋅(p1 −pQ ) + (ṗ 1 − ṗ Q )⋅(ṗ 1 − ṗ Q ) ] r∗1
is the (3 × 3) identity matrix. + [ p̈ 2 ⋅ (p2 −pQ ) + (ṗ 2 − ṗ Q )⋅(ṗ 2 − ṗ Q ) ] r∗2 + (61)
On the other hand, when the denominator of Eq. + [ p̈ 3 ⋅(p3 −pQ ) + (ṗ 3 − ṗ Q )⋅(ṗ 3 − ṗ Q ) ] r∗3 .
(58) equals zero, the author of [20] proposes an alter-
native formula:
ṗ Q = [ ṗ 1 ⋅ (p1 −pQ ) ] r∗1 + [ ṗ 2 ⋅ (p2 −pQ ) ] r∗2
𝜶 A2 = 2 J−1 vec (PP
̈ T − 𝛀2 R), for tr(P) = 0. (59) (62)
2 + [ ṗ 3 ⋅ (p3 −pQ ) ] r∗3
where

J = tr(R) I − R, ̇ T ≡ [q̇ ij ].
R ≡ PPT , PP
⎡ q̇ − q̇ 23 ⎤ ⎡ 𝜔1 ⎤ ⎡ 0 − 𝜔3 𝜔2 ⎤
1 ⎢ 32
̇ T) ≡
vec(PP q̇ − q̇ 31 ⎥, 𝝎A2 ≡⎢ 𝜔2 ⎥ = 2 J−1 vec(PP
̇ T ), 𝛀2 ≡ ⎢ 𝜔3 0 − 𝜔1 ⎥.
2 ⎢ 13 ⎥ ⎢ ⎥ ⎢ ⎥
⎣ q̇ 21 − q̇ 12 ⎦ ⎣ 𝜔3 ⎦ ⎣ −𝜔2 𝜔1 0 ⎦

and 𝛀22 ≡ 𝛀2 𝛀2. (p2 −pQ ) × (p3 −pQ ) (p3 −pQ ) × (p1 −pQ )
r∗1 = , r∗2 = ,
𝜎 𝜎
(p1 −pQ ) × (p2 −pQ )
4.5 Fifth representation of the angular acceleration r∗3 =
𝜎
.
vector (63)

A fifth representation of the angular acceleration vec- 𝜎 ≡ (p1 −pQ ) ⋅ [(p2 −pQ ) × (p3 −pQ )]. (64)
tor has been reported by Condurache and Matcovschi,
[13, 21], and the corresponding formulation is given and position vector pQ denotes the location of an
by: arbitrary point Q (a fourth point) of the moving rigid

( ){ }
𝜶 CM =
1
r∗1 × (p̈ 1 − p̈ Q ) + r∗2 × (p̈ 2 − p̈ Q ) + r∗3 × (p̈ 3 − p̈ Q ) (60)
2

where body which must be non-coplanar with points 1, 2,

and 3. Hence, we arbitrarily defined it as follows:

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 97

� �
1 � � (p −p ) × (p3 −p1 )
⎡ 0 ⎤ ⎡ −1 ⎤
pQ ≡ p1 +p2 +p3 + 2 1 ,
1⎢ ⎥
3 𝜒
ṗ = 0
√ ⎥, q = ⎢ 0 ⎥, ṗ ⋅ q = 0. (69)
4 ⎢√ ⎢ ⎥
⎣ 2− 3⎦ ⎣ 0 ⎦
√
𝜒≡ {(p2 −p1 ) × (p3 −p1 )} ⋅ {(p2 −p1 ) × (p3 −p1 )}, ∀ 𝜎 ≠ 0.
(65) and therefore, Eq. (40) fails to compute the angu-
lar velocity vector. However, this vector is required
Here it is important to note that the formulation of
to compute the corresponding angular acceleration
Condurache and Matcovschi, [13, 21], is the only
vector. Hence, Eq. (39) is not valid for the particular
one that explicitly involves a fourth point of the rigid
numerical data of this first case study.
body, which is not the centroid of the set of points 1,
2, and 3.
5.2 Angular acceleration of the second
representation
5 First case study
Equation (41) represents the angular acceleration vec-
The objective of this section is to show the applica- tor for the second representation. However, the com-
tion details of the different approaches to obtain the putation fails from the beginning for the numerical
angular acceleration vector. To this end, consider a data of this first case study since the denominator of
representative example taken from [20]. This exam- the corresponding angular velocity vector (45) van-
ple provides the position vectors, the velocity vectors, ishes, that is:
and the acceleration vectors of three non-collinear
points of a moving rigid body: ⎡ 0 ⎤ ⎡ −1 ⎤
1⎢ ⎥ ⎢ ⎥
vB∕A = ⎢ 0 ⎥, rC∕A = ⎢ 0 ⎥, vB∕A ⋅ rC∕A = 0.
1∕2 ⎤
⎡ √ ⎡ √0 ⎤ ⎡ −1∕2 ⎤ 4⎢ √ √
√ ⎥ ⎢ 0 ⎥
⎢ ⎥
p1 = − 3∕6 , p2 = ⎢ 3∕3 , p3 = ⎢ − 3∕6 ⎥,
⎥ ⎣ 2 − 3 ⎦ ⎣ ⎦
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ (70)
⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦
(66) and therefore, Eq. (45) fails to compute the angu-
lar velocity vector. However, this vector is required
√ ⎡0⎤ √ ⎡0⎤ √ ⎡0⎤
4 − 2⎢ ⎥ 4 − 3⎢ ⎥ 4 + 2⎢ ⎥ to compute the corresponding angular acceleration
4 ⎢⎢ ⎥⎥ 2 4 ⎢⎢ ⎥⎥ 3 4 ⎢⎢ ⎥⎥
ṗ 1 = 0 , p
̇ = 0 , p
̇ = 0 , vector. Hence, Eq. (41) is not valid for the particular
⎣1⎦ ⎣1⎦ ⎣1⎦ numerical data of this first case study.
(67)

√ √ √ √
⎡ −6 + 4 3 ⎤ ⎡8 3+3 6⎤ ⎡ 6+4 3 ⎤
1⎢ √ ⎥ 1⎢ √ ⎥ 1⎢ √ ⎥
p̈ 1 = ⎢ 12 − 3 2 ⎥, p̈ 2 = − ⎢ 3 3 ⎥, p̈ 3 = ⎢ −12 + 3 2 ⎥. (68)
24 ⎢ ⎥ 24 ⎢ ⎥ 24 ⎢ ⎥
⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦

The goal of this first case study is to obtain the angu- 5.3 Angular acceleration of the third representation
lar acceleration vector of the corresponding rigid
body using those approaches shown previously. For the angular acceleration vector of the third rep-
resentation, we have two choices, namely, Eqs. (47)
5.1 Angular acceleration of the first representation or (51). However, we need to compute the angular
velocity vector first. To this end, we observe that
Equation (39) is the angular acceleration vector of the denominator of the angular velocity vector 𝝎W1,
the first representation. The computation of the corre- Eq. (55), vanishes for the numerical data of this
sponding angular velocity vector (40) for the numeri- first case study. Hence, we resort to the alternative
cal data of this first case study fails from the begin- angular velocity vector 𝝎W2 , given by Eq. (56), thus
ning since the denominator vanishes, that is: yielding:

Vol.: (0123456789)
13
 98 Meccanica (2024) 59:89–106

⎡ −1 ⎤ thus resulting the following angular acceleration

1 ⎢√ ⎥ vector:
𝝎W2 = 2 . (71)
2⎢ ⎥
⎣ 0 ⎦
⎡0⎤
Now, since r3∕1 ⋅ q2∕3 ≠ 0, the angular acceleration 𝜶 A1 = ⎢ 0 ⎥ rad/s2 . (74)
⎢ ⎥
vector can be computed by the formula (47), which ⎣1⎦
yields the following numerical results:
√
⎡0⎤ ⎡− 3⎤ 5.5 Angular acceleration of the fifth representation
1
q1∕3 = ⎢ 1 ⎥, q2∕3 = ⎢ 1 ⎥,
⎢ ⎥ 2⎢ ⎥
⎣0⎦ ⎣ 0 ⎦ The corresponding angular acceleration vector is given
√ (72) by Eq. (60). In this case there are obtained the follow-
⎡ −1 ⎤
3 ing numerical results:
r3∕1 = ⎢ 0 ⎥, r3∕1 ⋅ q2∕3 = .
⎢ ⎥ 2
0 ⎡ 1
⎣ ⎦ ⎡0⎤ √ ⎤
3 ⎢ √ ⎥
pQ = ⎢ 0 ⎥, 𝜎=− , r∗1 = ⎢− 3 ⎥,
therefore, we see that the final result is given by: ⎢ ⎥ 2 3
⎣1⎦ ⎢ −1
⎣ 3
⎥
⎦
⎡0⎤
⎡ 0 ⎤ ⎡ −1 ⎤
𝜶 W1 = ⎢ 0 ⎥ rad/s2 . (73) ⎢ √ ⎥ ⎢ √ ⎥
⎢ ⎥
⎣1⎦ r∗2 = ⎢ 2 3 ⎥, r∗3 = ⎢− 3 ⎥,
3 3
⎢ −1 ⎥ ⎢ −1 ⎥
⎣ 3 ⎦ ⎣ 3 ⎦
√ √
⎡ 2 ⎤ ⎡− 6 ⎤
⎢ 21 ⎥ ⎢ √243 ⎥
5.4 Angular acceleration of the fourth representation ṗ Q = ⎢ 2 ⎥, p̈ Q = ⎢ − ⎥.
⎢ 24
√
⎢ 3 ⎥ 3 ⎥
The corresponding angular acceleration vector is ⎣ 1 − 12 ⎦ ⎣ −4 ⎦
given by Eq. (58). We readily obtain the following
which produce the following angular acceleration
numerical results for the particular data of the first
vector:
case study:

⎡ √
3 0 −3 ⎤ √ ⎡ 0 0 0 ⎤
1 √ √ 1 3 1 ⎢
P = ⎢ − 3 2 3 − 3 ⎥, tr(P) = + ,Ṗ = 0
√
0 0 ⎥,
6⎢ ⎥ 2 3 12 ⎢ √ √ √ ⎥
⎣ 0 0 0 ⎦ ⎣ 3−3 2 −2 3 3+3⎦
√ √ √ √ √ √
⎡ −1 ⎤ ⎡ −6 + 4 3 + 6 − 8 3 − 2 6 6 + 4 3 + 6 ⎤
1 ⎢√ ⎥ 1 ⎢ √ √ √ √ √ ⎥
𝝎A1 = 2 , P̈ = ⎢ 12 − 3 2 + 3 − 2 3 − 12 + 3 2 + 3 ⎥,
2⎢ ⎥ 24 ⎢ ⎥
⎣ 0 ⎦ ⎣ 0 0 0 ⎦
√ √
⎡ 6( 3+2)
√ 0 − 6 √3 ⎤ √
⎢ √ 3+2√3 √ 3+2
√ 3 ⎥ 24 3
12
D −1
= ⎢ − 2 3( √3+2) 4 2 3( √3−2) ⎥, 𝛼1 = √ , 𝛽=− √ .
⎢ 3+2 3 3+2 3 ⎥ 3+2 3 3+2 3
⎢ 0 0 12
√ ⎥
⎣ 3+2 3 ⎦

⎡0⎤
𝜶 CM = ⎢ 0 ⎥ rad/s2 . (75)
⎢ ⎥
⎣1⎦

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 99

5.6 Angular acceleration of the formula proposed in (b) The basic equations reported in [14] and [15]
this paper failed to compute the angular acceleration vector
corresponding to the numerical data of this first
Equation (36) represents the angular acceleration vec- case study.
tor proposed in this paper. The corresponding terms
for the numerical data of this first case study are the
following:

⎡ −1 ⎤ ⎡ −1 ⎤ √ √ √
1 ⎢√ ⎥ 3 2 3 3
L12 = 1, L13 = 1, u= 3 , m = ⎢ 0 ⎥, 𝜆= , k1 = , k2 = ,
2⎢ ⎥ ⎢ ⎥ 2 3 3
⎣ 0 ⎦ ⎣ 0 ⎦
√ √
⎡ 0 ⎤ ⎡ 0 ⎤ ⎡ 2− 6−4 3 ⎤ ⎡ 2 ⎤
1⎢ ⎥ 1⎢ √ √ ⎥ 1 ⎢√
u̇ = 0
√ ⎥, ṁ = ⎢ √ 0 ⎥, ü = ⎢ −4 + 2 − 3 ⎥, m ̈ = 2 − 4 ⎥,
4 ⎢√ ⎢ ⎥ 8⎢ ⎥ 4⎢ ⎥
⎣ 2− 3⎦ ⎣ 2∕2 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦

⎡0⎤ √ ⎡ 0 ⎤ ⎡0⎤ √ ⎡ 0 ⎤
3⎢ ⎥, 𝜿 3 = ⎢ 0 ⎥, 𝜿 4 = 6 ⎢
𝜿 1 = ⎢ 0 ⎥, 𝜿2 = 0 0 ⎥
√ ⎥,
⎢ ⎥ 16 ⎢ √ √ ⎥ ⎢ ⎥ 16 ⎢ √
⎣0⎦ ⎣ 6+4 3−2⎦ ⎣ ⎦
0 ⎣ 2− 3⎦
⎡0⎤ ⎡0⎤ ⎡0⎤ √ ⎡ 0 ⎤ √ ⎡ 0 ⎤
3⎢ ⎥, 𝜿 = 3 ⎢
𝜿 5 = ⎢ 0 ⎥, 𝜿 6 = ⎢ 0 ⎥, 𝜿 7 = ⎢ 0 ⎥, 𝜿 8 = 0 0
√ ⎥.
⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ 32 ⎢ √ ⎥ 9 32 ⎢ √
⎣0⎦ ⎣ ⎦
0 ⎣ ⎦
0 ⎣2 6−5⎦ ⎣ ( 3 − 2) ⎦ 2

which produce the following angular acceleration 6 Second case study

vector:
The second case study considers a very representa-
⎡0⎤ tive example taken from [15]. This example pro-
𝜶 = ⎢ 0 ⎥ rad/s2 . (76) vides the position vectors, the velocity vectors,
⎢ ⎥
⎣1⎦ and the acceleration vectors of three non-collinear
points of a moving rigid body, namely:

⎡ 100 ⎤ ⎡ 300 ⎤ ⎡ 220 ⎤

5.7 Remarks on the first case study p1 = ⎢ 100 ⎥, p2 = ⎢ 300 ⎥, p3 = ⎢ 180 ⎥, (77)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦
A careful analysis of the numerical results related to
previous computations shows that:
⎡ 600 ⎤ ⎡ 200 ⎤ ⎡ 440 ⎤
(a) The formula proposed in this paper and the for- ṗ 1 = ⎢ −400 ⎥, ṗ 2 = ⎢ 0 ⎥, ṗ 3 = ⎢ −160 ⎥, (78)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
mulas that appear in [13, 17], and [20] were ⎣ 100 ⎦ ⎣ 0 ⎦ ⎣ 40 ⎦
able to compute the correct numerical value
of the angular acceleration vector correspond- ⎡ 850 ⎤ ⎡ 200 ⎤ ⎡ 420 ⎤
ing to the numerical data of this first case study. ⎢ ⎥ ⎢ ⎥
p̈ 1 = 1200 , p̈ 2 = 200 , p̈ 3 = ⎢ 760 ⎥. (79)
All the formulas produced the same numerical ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ −240 ⎦ ⎣ 0 ⎦ ⎣ −140 ⎦
result.
which are given in millimeters, millimeters per sec-
ond, and millimeters per second squared, respectively.

Vol.: (0123456789)
13
 100 Meccanica (2024) 59:89–106

The objective of this second case study is to obtain 6.3 Angular acceleration of the third representation
the angular acceleration vector of the moving rigid
body using all five formulas shown previously. The angular acceleration vector for the third repre-
sentation is given by Eq. (47). The numerical values
associated with this formula are the following:
6.1 Angular acceleration of the first representation
⎡ 0 ⎤
Equation (39) represents the angular accelera- 𝝎W1 = ⎢ 1∕2 ⎥. (82)
tion vector of the first representation. This formula ⎢ ⎥
⎣ 2 ⎦
requires the following numerical computations:

⎡ 200 ⎤ ⎡ 120 ⎤ ⎡ −400 ⎤ ⎡ −160 ⎤ ⎡ −650 ⎤

p = ⎢ 200 ⎥, q = ⎢ 80 ⎥, ṗ = ⎢ 400 ⎥, q̇ = ⎢ 240 ⎥, p̈ = ⎢ −1000 ⎥,
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ −100 ⎦ ⎣ −60 ⎦ ⎣ 240 ⎦
⎡ −430 ⎤ ⎡ 0 ⎤ ⎡ 1∕4 ⎤ ⎡ 3∕20 ⎤
q̈ = ⎢ −440 ⎥, 𝝎FZ = ⎢ 1∕2 ⎥, s1 = ⎢ 1∕4 ⎥, s2 = ⎢ 1∕10 ⎥,
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ 100 ⎦ ⎣ 2 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦
⎡ 0 ⎤ ⎡ 0 ⎤ ⎡ 0 ⎤
s3 = ⎢ −1∕8 ⎥, s4 = ⎢ −85∕16 ⎥, s5 = ⎢ 87∕16 ⎥.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ −1∕2 ⎦ ⎣ −85∕4 ⎦ ⎣ 87∕4 ⎦

which produce the following numerical result for the Now, since r3∕1 ⋅ q2∕3 ≠ 0, the angular acceleration
angular acceleration vector: vector can be computed by the formula (47), which
yields the following numerical results:
⎡ 1∕10 ⎤
𝜶 FZ = ⎢ −1∕10 ⎥ rad/s2 . (80) ⎡ −80 ⎤ ⎡ 120 ⎤
⎢ ⎥
⎣ −1 ⎦ q1∕3 = ⎢ 120 ⎥, q2∕3 = ⎢ −80 ⎥,
⎢ ⎥ ⎢ ⎥
⎣ −20 ⎦ ⎣ 20 ⎦
(83)
6.2 Angular acceleration of the second ⎡ 120 ⎤
representation r3∕1 = ⎢ 80 ⎥, r3∕1 ⋅ q2∕3 = 8000.
⎢ ⎥
⎣ 0 ⎦
The corresponding angular acceleration vector is
given by Eq. (41). The numerical values associated thus, we see that the final result is given by:
with this formula are the following:
⎡ 1∕10 ⎤
⎡ −400 ⎤ ⎡ −160 ⎤ ⎡ 120 ⎤ 𝜶 W1 = ⎢ −1∕10 ⎥ rad/s2 . (84)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
vB∕A =⎢ 400 ⎥, vC∕A = ⎢ 240 ⎥, rC∕A = ⎢ 80 ⎥, ⎣ −1 ⎦
√
⎢ 2 − 100 ⎥ ⎢ −60 ⎥ ⎢ 0 ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
⎡ 0 ⎤ ⎡ 200 ⎤ ⎡ 80 ⎤ 6.4 Angular acceleration of the fourth representation
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
𝝎SLI = ⎢ 1∕2 ⎥, qB∕A = ⎢ −200 ⎥, qC∕A = ⎢ −120 ⎥.
⎢ 2 ⎥ ⎢ 40 ⎥ ⎢ 20 ⎥ Equation (58) is the basic formula for the angular accel-
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
eration vector of the fourth representation. A quick cal-
and we get the following angular acceleration vector: culation of the matrix P for the particular data of the
second case study reveals that:
⎡ 1∕10 ⎤
𝜶 SLI = ⎢ −1∕10 ⎥ rad/s2 . (81)
⎢ ⎥
⎣ −1 ⎦

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 101

⎡ −320 280 40 ⎤ which produce the following angular acceleration

1
P = ⎢ −280 320 − 40 ⎥, tr (P) = 0. (85) vector:
3⎢ ⎥
⎣ 0 0 0 ⎦
⎡ 1∕10 ⎤
but the term tr(P) is part of the denominator of the 𝜶 CM = ⎢ −1∕10 ⎥ rad/s2 . (87)
⎢ ⎥
parameters 𝛼1, and 𝛽 , which are related to the angular ⎣ −1 ⎦
acceleration vector given by Eq. (58). Therefore, we
have to resort to the alternative formula of the angular
acceleration vector (59), which produces the follow-
ing numerical results:

⎡ 60800 59200 0 ⎤ ⎡ 19 37
0 ⎤
1⎢ 121600 ⎢ 20000 40000 ⎥
R= 59200 60800 0 ⎥, tr (R) = , J −1 37
= ⎢ 40000 19
0 ⎥,
3⎢ ⎥ 3 ⎢ 0
20000
⎥
⎣ 0 0 0⎦
⎣ 0 3
⎦
121600

⎡ −118400 − 121600 0 ⎤ ⎡ 0 ⎤ ⎡ 0 −4 1⎤
̇ T = 1 ⎢ 121600 118400 0 ⎥,
PP 𝝎A2 = ⎢ 1∕2 ⎥, 𝛀2 =
1⎢
4 0 0 ⎥.
3⎢ ⎥ ⎢ ⎥ 2⎢ ⎥
⎣ −30400 − 29600 0 ⎦ ⎣ 2 ⎦ ⎣ −1 0 0 ⎦

and, finally, it is obtained the following angular accel- 6.6 Angular acceleration of the formula proposed in
eration vector: this paper
⎡ 1∕10 ⎤
Equation (36) represents the angular acceleration vec-
𝜶 A2 = ⎢ −1∕10 ⎥ rad/s2 . (86)
⎢ ⎥ tor proposed in this paper. The corresponding computa-
⎣ −1 ⎦
tions for the numerical data of the second case study are
the following:
6.5 Angular acceleration of the fifth representation

The corresponding angular acceleration vector is given

by Eq. (60). In this particular case, the following results
are obtained:

620
⎡ 3 ⎤ ⎡− 3 ⎤ ⎡− 1 ⎤ ⎡ 1 ⎤
⎢ 200 ⎥ ⎢ 100 ⎥ ⎢ 401 ⎥
pQ = ⎢ 580 ⎥, 𝜎 = −8000, 1
r∗1 = ⎢ 100 ⎥,
3
r∗2 = ⎢ 200 ⎥, r∗3 = ⎢ − 40 ⎥,
⎢ 3 ⎥ ⎢ 1 ⎥ ⎢ 1 ⎥ ⎢ 1 ⎥
⎣ −1 ⎦ ⎣ 3 ⎦ ⎣ 3 ⎦ ⎣ 3 ⎦
⎡ 2477 ⎤ ⎡ 4901 ⎤
⎢ 6 ⎥ ⎢ 10 ⎥
ṗ Q = ⎢ − 560
3 ⎥, p̈ Q = ⎢ 7191
10 ⎥.
⎢ 140 ⎥ ⎢ − 1517 ⎥
⎣ 3 ⎦ ⎣ 12 ⎦

Vol.: (0123456789)
13
 102 Meccanica (2024) 59:89–106

√ ⎡1⎤ √ ⎡3⎤
√ √ 2⎢ ⎥ 13 ⎢ ⎥
L12 = 200 2 mm, L13 = 40 13 mm, u= 1 , m= 2 ,
2 ⎢ ⎥ 13 ⎢ ⎥
⎣0⎦ ⎣0⎦
√ √ ⎡ −4 ⎤ √ ⎡ −8 ⎤
26 √ 2⎢ 13 ⎢
𝜆= , k1 = 26, k2 = 5, u̇ = 4 ⎥, ṁ = 12 ⎥,
26 4 ⎢ ⎥ 26 ⎢ ⎥
⎣ −1 ⎦ ⎣ −3 ⎦
√ ⎡ −65 ⎤ √ ⎡ −43 ⎤ ⎡ −5 ⎤
2⎢ 13 ⎢ 1 ⎢
ü = −100 ⎥, ̈ =
m −44 ⎥, 𝜿 1 = −5 ⎥,
40 ⎢ ⎥ 52 ⎢ ⎥ 52 ⎢ ⎥
⎣ 24 ⎦ ⎣ 10 ⎦ ⎣ 0 ⎦
⎡ 9 ⎤ ⎡1⎤ ⎡ −12 ⎤
⎢ 65 ⎥ 1 ⎢ ⎥ 1 ⎢
𝜿2 = ⎢ 6
⎥, 𝜿3 = 1 , 𝜿4 = 12 ⎥,
⎢
65
⎥ 13 ⎢ ⎥ 104 ⎢ ⎥
⎣
79
⎦ ⎣0⎦ ⎣ −3 ⎦
104

⎡ 2 ⎤ ⎡ −2 ⎤ ⎡ 0 ⎤
1 ⎢ 1 ⎢ 1 ⎢
𝜿5 = −3 ⎥, 𝜿 6 = 3 ⎥, 𝜿 7 = −5 ⎥,
26 ⎢ ⎥ 26 ⎢ ⎥ 26 ⎢ ⎥
⎣ −20 ⎦ ⎣ 20 ⎦ ⎣ −20 ⎦
√ ⎡0⎤ √ ⎡ 0 ⎤
33 26 ⎢ ⎥ 33 26 ⎢
𝜿8 = 0 , 𝜿9 = 0 ⎥.
208 ⎢ ⎥ 208 ⎢ ⎥
⎣1⎦ ⎣ −1 ⎦

which produce the following angular acceleration 7 Conclusions

vector:
This paper presented a systematic and detailed
⎡ 1∕10 ⎤
approach for the computation of the angular accelera-
𝜶 = ⎢ −1∕10 ⎥ rad/s2 . (88)
⎢ ⎥ tion vector of a rigid body moving in space. Moreo-
⎣ −1 ⎦
ver, the proposed approach systematically groups
some desirable features, which leads to a simpler and
better way of conceiving and understanding the cor-
6.7 Remarks on the second case study
relation that exists between the rotational motion of
a rigid body and the velocity and acceleration of an
A detailed analysis of all the numerical results related
arbitrary point of the body. Some advantages of the
to the previous computations shows that:
proposed approach are the following:
(a) The equations reported in [13–15, 17], and the
(1) The novelty of the proposed method lies in the
formula proposed in this paper were able to
particular form of writing the pose of the moving
compute the angular acceleration vector cor-
rigid body, as well as in the procedure to com-
responding to the numerical data of the second
pute its time derivatives. As a result, the method
case study. All the formulas produced the same
leads naturally to obtaining a novel expression of
numerical result.
the angular acceleration vector, where it is not
(b) Only the basic formula proposed in [20] failed
required to solve the inverse acceleration problem
to compute the angular acceleration vector cor-
in rigid body kinematics [13].
responding to the numerical data related to the
(2) The integrated nature of the proposed approach
second case study. However, the alternative for-
may help to visualize the physical connection
mula shown in [20] produces the correct numeri-
between rotation, angular velocity, and angular
cal value of the angular acceleration vector.

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 103

acceleration, which are present during the spa- Appendix 1

tial motion of a rigid body. On the one hand, the
proposed method treats the rotational phenom- The objective of this appendix is to present a detailed
enon in a way that agrees directly with physical derivation of the nine vector terms ṅ 1 , ṅ 2 , ⋯ ṅ 9
insight. On the other hand, the approach allows involved into Eq. (30). To this end, the general idea is
us to correlate the angular motion of the body to include only u, u̇ , ü , m, ṁ , and m
̈ , since these vec-
with the translational motion of any point of the tors are directly related to the position, the velocity,
moving body. and the acceleration of the three non-collinear points
(3) The approach presented in this paper does not of the moving rigid body under analysis.
require the computation of the angular veloc-
ity vector of the moving body. However, all the (1) Computation of the first term, ṅ 1.
representations of the angular acceleration vec-
tor reported in [14, 15, 17], and [20] requires the The first term, namely, ṅ 1, has been previously
computation of the angular velocity vector. This defined in Eq. (30), and it may be handled as follows:
issue may represent a serious drawback since
ṅ 1 ≡ (v⋅w)u
̈ ̈ − k2 u)
= {(k1 m ̈ ⋅ (u × v)}u
obtaining the angular velocity vector may lead to (89)
computational singularities. = k1 {m
̈ ⋅ (u × v)}u − k2 {ü ⋅ (u × v)}u.
(4) A singularity occurs when the denominator of
By using vector product identities, we have that:
the angular acceleration vector equals zero. The
denominators of the angular acceleration vec- (u×v)×(u× m)
̈ = {m⋅(u×v)}u
̈ − {u⋅(u×v)}m
̈
tors reported in [14, 15, 17], and [20] involves (90)
= {m⋅(u×v)}u,
̈ u⋅(u×v) = 0.
several vectors, in consequence, they are sensi-
tive to singularities. That is because these results
(u×v)×(u× u)
̈ = {ü ⋅ (u×v)}u − {u ⋅ (u×v)}ü
come from an intensive vector and matrix algebra (91)
manipulation of preconceived equations of clas- = {ü ⋅ (u×v)}u, u ⋅ (u×v) = 0.
sical kinematics. On the other hand, the denomi- where
nator of the expression for the angular accelera-
tion vector proposed in this paper is very simple, u×v = u×(k1 m − k2 u) = k1 (u×m)
namely, the scalar parameter 𝜆2, and therefore, it (92)
− k2 (u×u) = k1 (u×m), u×u = 0.
is very robust against singularities. It may pro-
duce a singularity if, and only if, the three points In this way, according to the rules of cross vector
of the body are collinear. products, Eq. (89) may be written as follows:

Last, the authors hope that the present contribution

ṅ 1 = k12 (m×u)×(u×m)
̈ − k1 k2 (u×u)×(u×m).
̈ (93)
may help to a better understanding of the accelera-
tion analysis of complex multibody systems, such (2) Computation of the second term, ṅ 2.
as machines, mechanisms, parallel robots, and the
human body. The algebraic handling of the second term, namely,
ṅ 2, is described below:
Acknowledgements The authors are very thankful to the
National Council of Humanities, Science and Technology of ṅ 2 ≡ (v⋅ ̇ = {(k1 ṁ − k2 u)
̇ w)u ̇ ⋅ (u̇ × v + u × v)}u
̇
México (CONAHCYT) for supporting this research through
SNI (National System of Researchers) fellowships.
(94)
where
Declarations
u̇ × v = u̇ × (k1 m − k2 u) = k1 (u̇ × m) − k2 (u̇ × u)
Conflicts of interest The authors declare that they have no (95)
conflict of interest.
̇ = k1 (u × m)
u × v̇ = u × (k1 ṁ − k2 u) ̇ − k2 (u × u)
̇
(96)
Thus, vector ṅ 2 becomes:

Vol.: (0123456789)
13
 104 Meccanica (2024) 59:89–106

ṅ 2 = k12 {ṁ ⋅ (u̇ × m)}u − k1 k2 {u̇ ⋅ (u × m)}u

̇ (97) ṅ 4 = k12 (u×m)×(u×m)
̈ − k1 k2 (u×u)×(u×m).
̈ (106)

Next, by resorting to the following vector product (5) Computation of the fifth term, ṅ 5.
identity, it is found that: The fifth term can be formulated in a way that
yields a convenient vector expression, which starts
(u×u)×(u×
̇ ̇ = {u⋅(u×
m) ̇ m)}u−{u⋅(u×
̇ m)}
̇ u̇
(98) with the relation:
= {u⋅(u×
̇ m)}u,
̇ ̇ = 0.
u⋅(u× m)
ṅ 5 ≡ (u⋅ ̇ = {u̇ ⋅ (u̇ × v + u × v)}
̇ w)v ̇
Therefore, it is finally obtained that: v = {u̇ ⋅ (u̇ × v) + u̇ ⋅ (u × v)}v
̇ (107)
ṅ 2 = k12 {ṁ ⋅ (u̇ × m)}u − k1 k2 (u×u)×(u×
̇ m).
̇ (99) ṅ 5 = {u̇ ⋅ (u × v)}v,
̇ u̇ ⋅ (u̇ × v) = 0.

where
(3) Computation of the third term, ṅ 3. ̇ = k1 (u × m)
u × v̇ = u × (k1 ṁ − k2 u) ̇ − k2 (u × u)
̇
(108)
This section shows the computation of the third term,
namely, ṅ 3. The process begins with the following u⋅(u×
̇ ̇ = k1 {u⋅(u×
v) ̇ m)}−k
̇ 2 {u⋅(u×
̇ u)}
̇
(109)
expression: = k1 {u⋅(u×
̇ m)},
̇ u⋅(u×
̇ ̇ = 0.
u)

ṅ 3 ≡ (v⋅w)
̇ u̇ = {(k1 ṁ − k2 u)
̇ ⋅ (u × v)}u̇ (100) Then Eq. (107) becomes:

where ṅ 5 = k1 {u̇ ⋅ (u× m)}(k

̇ 1 m − k2 u)
(110)
u×v = u×(k1 m − k2 u) = k1 (u×m) = k12 {u̇ ⋅ (u × m)}m
̇ − k1 k2 {u̇ ⋅ (u × m)}u.
̇
(101)
− k2 (u×u) = k1 (u×m), u×u = 0. To complete the reduction process, we now use the
well-known vector product identity:
Then, the third term can be expressed as:
(u×u)×(u×
̇ ̇ = {u⋅(u×
m) ̇ m)}u
̇ − {u⋅(u× m)}
̇
̇ (102)
ṅ 3 = k12 {ṁ ⋅ (u × m)}u̇ − k1 k2 {u̇ ⋅ (u × m)}u. (111)
u̇ = {u⋅(u×
̇ m)}u,
̇ ̇ = 0.
u⋅(u× m)

Thus, the sought expression is, therefore:

(4) Computation of the fourth term, ṅ 4.
ṅ 5 = k12 {u̇ ⋅ (u × m)}m
̇ − k1 k2 (u×u)×(u×
̇ m).
̇ (112)
The algebraic manipulation of the fourth term is as
follows:
(6) Computation of the sixth term, ṅ 6.
ṅ 4 ≡ (u⋅w)v
̈ = {ü ⋅ (u × v)}v = {ü ⋅ (k1 u × m)}(k1 m − k2 u)
ṅ 4 = k12 {ü ⋅ (u × m)}m − k1 k2 {ü ⋅ (u × m)}u. In this section, we examine another means of
(103) expressing the so-called sixth term. To this end, in the
Recalling the following vector product identities: first instance, we have the following equation:

(u×m)×(u×m)
̈ = {u⋅(u×m)}m−{m⋅(u×m)}
̈ ü = {u⋅(u×m)}m,
̈ m⋅(u×m) = 0. (104)

(u×u)×(u×m)
̈ = {u⋅(u×m)}u−{u⋅(u×m)}
̈ ü = {u⋅(u×m)}u,
̈ u⋅(u×m) = 0. (105)

Having laid the necessary groundwork, we get the

following result:

Vol:. (1234567890)
13
Meccanica (2024) 59:89–106 105

ṅ 6 ≡ (u⋅w)
̇ v̇ = {u⋅(u
̇ × v)}v̇ = {u⋅(k
̇ 1 u × m)}v̇ (9) Computation of the ninth term, ṅ 9.
= {u⋅(k
̇ 1 u × m)}(k1 ṁ − k2 u)
̇ This section is dedicated to finding an alternative
(113) formula for the ninth term defined in Eq. (30). This
and, after some vector algebra, we obtain the final term can be expressed as:
result given by:
ṅ 9 ≡ (u⋅v)
̇ ẇ = {u⋅(k
̇ 1 m − k2 u)}ẇ = {k1 (u⋅
̇ m) − k2 (u⋅
̇ u)}(u̇ × v + u × v)
̇
(121)
̇ (114)
ṅ 6 = k12 {u̇ ⋅ (u × m)}ṁ − k1 k2 {u̇ ⋅ (u × m)}u. where
u̇ × v = u̇ × (k1 m − k2 u) = k1 (u̇ × m) − k2 (u̇ × u)
(7) Computation of the seventh term, ṅ 7. (122)
In this section, we analyze the mathematical form ̇ = k1 (u × m)
u × v̇ = u × (k1 ṁ − k2 u) ̇ − k2 (u × u)
̇
associated with the seventh term described in Eq. (123)
(30), namely:

ṅ 7 ≡ (u⋅v)w
̈ = {u⋅(k
̈ 1 m − k2 u)}w = {u⋅(k
̈ 1 m − k2 u)}(u × v)
(115)
̈ 1 m − k2 u)}(k1 u × m) = k12 (ü ⋅ m)(u × m)−k1 k2 (ü ⋅ u)(u × m)
ṅ 7 = {u⋅(k

The first term of the above equation may be conveni-

̇ u = 0,
u⋅ u̇ × u = −u × u.
̇ (124)
ently transformed by the following vector identity:
From these relationships, we have that:
ü × {(u × m) × m} = (ü ⋅ m)(u × m) − {ü ⋅ (u × m)}m
(116) ṅ 9 = k12 (u̇ ⋅ m)(u̇ × m) + k12 (u̇ ⋅ m)(u × m).
̇ (125)
Thus, the seventh term is given by:
The algebraic process continues using the following
ṅ 7 = k12 ü × {(u × m) × m} two vector identities:
(117)
+ k12 {ü ⋅ (u × m)}m − k1 k2 (ü ⋅ u)(u × m). u×{(
̇ u×m)×m}
̇ = (u⋅m)(
̇ u×m)−{
̇ u⋅(
̇ u×m)}m
̇
= (u⋅m)(
̇ u×m),
̇ u⋅(
̇ u×m)
̇ = 0.
(8) Computation of the eighth term, ṅ 8. (126)
u×{(u×
̇ m)×m}
̇ = (u⋅m)(u×
̇ m)−{
̇ u⋅(u×
̇ m)}m.
̇
We now consider an alternative derivation of (127)
the formula for the eighth term that was previously By using the foregoing identities it is obtained the
defined. The procedure is as follows: final result given by:

ṅ 8 ≡ (u⋅
̇ v)w
̇ = {u⋅(k
̇ 1 ṁ − k2 u)}w
̇ = {u⋅(k
̇ 1 ṁ − k2 u)}(u
̇ × v)
2
(118)
ṅ 8 = {u⋅(k
̇ 1 m−k
̇ 2 u)}(k
̇ 1 u×m) = k1 (u̇ ⋅ m)(u
̇ × m)−k1 k2 (u⋅
̇ u)(u×m)
̇

To get a more convenient form of the first term of ṅ 9 = k12 u×{(

̇ u×m)×m}
̇ + k12 u×{(u×
̇ m)×m}
̇ + k12 {u⋅(u×
̇ m)}m.
̇
the above equation, we resort to the following vector (128)
identity:
u̇ × {(u × m) × m}
̇ = (u̇ ⋅ m)(u
̇ × m) − {u̇ ⋅ (u × m)}ṁ References
(119)
1. Euler L (1775) Nova methodus motum corporum rigido-
Therefore, the eighth term gets the following form: rum determinandi. Novi Commentari Acad Imp Petrop
20:208–238
ṅ 8 = k12 u̇ × {(u × m) × m}
̇ + k12 {u̇ ⋅ (u × m)}ṁ − k1 k2 (u̇ ⋅ u)(u
̇ × m). 2. Rodrigues O (1840) Des lois géométriques qui régissent
(120) les déplacements d’un systéme solide dans l’espace, et de

Vol.: (0123456789)
13
 106 Meccanica (2024) 59:89–106

la variation des coordonnées provenant de ces déplace- 14. Field P, Ziwet A (1916) The accelerations of the points of
ments considérés indépendamment des causes qui peuvent a rigid body. Am Math Mon 23(10):371–381
les produir. J Math 5:380–440 15. Soutas-Little RW, Inman DJ (1999) Engineering mechan-
3. Shuster MD (1993) A survey of attitude representations. J ics: dynamics. Prentice-Hall Inc, USA, pp 386–387
Astron Sci 41(4):439–517 16. Wittenburg J (2008) Dynamics of multibody systems, 2nd
4. Trindade MA, Sampaio R (2000) Uma Revisão sobre a edn. Springer, Cambridge, pp 23–26
Parametrização de Rotações Finitas na Dinâmica de Cor- 17. Wittenburg J (2016) Kinematics. Springer, UK, pp
pos Rígidos. J Braz Soc Mech Sci 222(2):341–377 316–317
5. Caparrini S (2002) The discovery of the vector represen- 18. Angeles J (1987) Computation of rigid-body angu-
tation of moments and angular velocity. Arch Hist Exact lar acceleration from point-acceleration measurements.
Sci 56:151–181 ASME J Dyn Syst Meas Control 109:124–127
6. Condurache D (2022) Higher-order relative kinematics of 19. Angeles J (1999) The angular-acceleration tensor of rigid-
rigid body and multibody systems. A novel approach with body kinematics and its properties. Arch Appl Mech
real and dual lie algebras. Mech Mach Theory 176:1–27 69:204–214
7. Mohamed MG (1997) Kinematics of rigid bodies in gen- 20. Angeles J (2014) Fundamentals of robotic mechanical
eral spatial motion: second-order motion properties. Appl systems, 4th edn. Springer, New York, pp 362–368
Math Model 21(8):471–479 21. Condurache D, Matcovschi M (2002) Computation of
8. McGill DJ, King WW (1984) Engineering mechanics: an angular velocity and acceleration tensors by direct meas-
introduction to dynamics, Brooks/Cole Engineering Divi- urements. Acta Mech 153:147–167
sion, pp 490–492 22. Cervantes-Sánchez JJ, Rico-Martínez JM, Pérez-Muñoz
9. Hassani A, Khalilpour SA, Bataleblu A, Taghirad HD VH (2014) On the angular velocity of a rigid body:
(2022) Full dynamic model of 3-UPU translational paral- matrix and vector representations. Eur J Mech A/Solids
lel manipulator for model-based control schemes. Robot- 45:123–132
ica 40(8):2815–2830 23. Rico-Martínez JM, Duffy J (1996) An application of
10. Lewkowiciz R, Kowaleczko G (2019) An inverse kin- screw algebra to the acceleration analysis of serial chains.
ematic model of the human training centrifuge motion Mech Mach Theory 31(4):445–457
simulator. J Theor Appl Mech 57(1):99–113
11. Matuszek DB, Bilos LA (2021) Computer image analy- Publisher’s Note Springer Nature remains neutral with regard
sis as a method of evaluating the quality of selected fine- to jurisdictional claims in published maps and institutional
grained food mixtures. Sustainability 13(3018):1–14 affiliations.
12. Carlsen RW, Fawzi AL, Wan F, Kesari H, Franck C
(2021) A quatitative relationship between rotational head
Springer Nature or its licensor (e.g. a society or other partner)
kinematics and brain tissue strain from a 2-D parametric
holds exclusive rights to this article under a publishing
finite element analysis. Brain Multiphys 2(100024):1–14
agreement with the author(s) or other rightsholder(s); author
13. Condurache D, Matcovschi MH (2001) Algebraic compu-
self-archiving of the accepted manuscript version of this article
tation of the twist of a rigid body through direct measure-
is solely governed by the terms of such publishing agreement
ments. Comput Methods Appl Mech Eng 190:5357–5376
and applicable law.

Vol:. (1234567890)
13