arXiv:1104.1249v1  [cs.RO]  7 Apr 2011
The Design of a Novel Prismatic Drive
for a Three-DOF Parallel-Kinematics Machine
D. Chablat1, J. Angeles2
1Institut de Recherche en Communications et Cybern´etique de Nantes ∗
UMR CNRS n◦6597, 1 rue de la No¨e, 44321 Nantes, France
2Department of Mechanical Engineering &
Centre for Intelligent Machines, McGill University
817 Sherbrooke Street West, Montreal, Canada H3A 2K6
Damien.Chablat@irccyn.ec-nantes.fr
angeles@cim.mcgill.ca
November 7, 2018
Abstract
The design of a novel prismatic drive is reported in this paper. This transmission is based on Slide-o-
Cam, a cam mechanism with multiple rollers mounted on a common translating follower. The design of
Slide-o-Cam was reported elsewhere. This drive thus provides pure-rolling motion, thereby reducing the
friction of rack-and-pinions and linear drives.
Such properties can be used to design new transmissions
for parallel-kinematics machines. In this paper, this transmission is intended to replace the ball-screws in
Orthoglide, a three-dof parallel robot intended for machining applications.
1
Introduction
In robotics and mechatronics applications, whereby motion is controlled using a piece of software, the conversion
of motion from rotational to translational is usually done by either ball-screws or linear actuators. Of these
alternatives, ball-screws are gaining popularity, one of their drawbacks being the high number of moving parts
that they comprise, for their functioning relies on a number of balls rolling on grooves machined on a shaft; one
more drawback of ball-screws is their low load-carrying capacity, stemming from the punctual form of contact
by means of which loads are transmitted. Linear bearings solve these drawbacks to some extent, for they can be
fabricated with roller-bearings; however these devices rely on a form of direct-drive motor, which makes them
expensive to produce and to maintain.
A novel transmission, called Slide-o-Cam, was introduced in [1] as depicted in Fig. 1, to transform a rotation
into a translation. Slide-o-Cam is composed of four major elements: (i) the frame, (ii) the cam, (iii) the
follower, and (iv) the rollers. The input axis on which the cam is mounted, the camshaft, is driven at a constant
angular velocity by an actuator under computer-control. Power is transmitted to the output, the translating
follower, which is the roller-carrying slider, by means of pure-rolling contact between cam and roller. The roller
comprises two components, the pin and the bearing. The bearing is mounted at one end of the pin, while the
other end is press-ﬁt into the roller-carrying slider. Contact between cam and roller thus takes place at the
outer surface of the bearing. The mechanism uses two conjugate cam-follower pairs, which alternately take
over the motion transmission to ensure a positive action; the rollers are thus driven by the cams throughout a
complete cycle. The main advantage of cam-follower mechanisms over alternative transmissions to transform
rotation into translation is that contact through a roller reduces friction, contact stress and wear.
This transmission, once fully optimized, will replace the three ball-screws used by the Orthoglide prototype
[2]. Orthoglide features three prismatic joints mounted orthogonaly, three identical legs and a mobile platform,
which moves in Cartesian space with ﬁxed orientation, as shown in Fig. 2. The three motors are SANYO
∗IRCCyN: UMR n◦6597 CNRS, ´Ecole Centrale de Nantes, Universit´e de Nantes, ´Ecole des Mines de Nantes
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
2
DENKI (ref. P30B08075D) with a constant torque of 1.2 Nm from 0 to 3000 rpm. This property enables the
mechanism to move throughout the workspace a 4 kg load with an acceleration of 17 ms−2 and a velocity of
1.3 ms−1. On the ball-screws, the pitch is 50 mm per cam turn. The minimum radius of the camshaft is 8.5 mm.
A new arrangement of camshaft, rollers and follower is proposed to reduce the inertial load when more than
two cams are used.
PSfrag replacements
Roller
Follower
Conjugate cams
Figure 1: Layout of Slide-o-Cam
Figure 2: The Orthoglide
Unlike Lampinen [3], who used a genetic algorithm, or Zhang and Shin [4], who used what they called a
relative-motion method, where the relative velocity and relative acceleration of a follower with respect to a cam
is prescribed and implemented in C++, we use a deterministic method while taking into account geometric
and machining constraints, as outlined in [5]. Unlike Carra [6], who used a negative-radius follower to reduce
the pressure angle, we use a positive-radius follower that permits us to assemble several cams on the same
roller-carrier.
In Section 2, we introduce the relations describing the cam proﬁle and the mechanism kinematics.
In
Section 3, we derive conditions on the design parameters so as to have a fully convex cam proﬁle, to avoid
undercutting, and to have a geometrically feasible mechanism. In Section 4, the pressure angle, a key performance
index of cam mechanisms, is studied in order to choose the design parameters that give the best pressure angle
distribution, a compromise being made with the accuracy of the mechanism.
2
Synthesis of the Planar Cam Mechanism
Let the x-y frame be ﬁxed to the machine frame and the u-v frame be attached to the cam, as depicted in
Fig. 3. O1 is the origin of both frames, while O2 is the center of the roller, and C is the contact point between
cam and roller. The geometric parameters deﬁning the cam mechanism are illustrated in the same ﬁgure. The
notation of this ﬁgure is based on the general notation introduced in [7], namely, (i) p, the pitch, i.e., the
distance between the center of two rollers on the same side of the follower; (ii) e, distance between the axis of
the cam and the line of centers of the rollers; (iii) a4, radius of the roller-bearing, i.e., the radius of the roller;
(iv) ψ, angle of rotation of the cam, the input of the mechanism; (v) s, position of the center of the roller, i.e,
the displacement of the follower, the output of the mechanism; (vi) µ, pressure angle; (vii) f, force transmitted
from cam to roller. In this paper, p is set to 50 mm, in order to meet the Orthoglide speciﬁcations.
The above parameters as well as the contact surface on the cam, are determined by the geometric relations
dictated by the Aronhold-Kennedy Theorem in the plane [8]. When the cam makes a complete turn (∆ψ = 2π),
the displacement of the roller is equal to p, the distance between two rollers on the same side of the roller-
carrying slider (∆s = p). Furthermore, if we consider the home conﬁguration of the roller as depicted in Fig. 4,
the roller is below the x-axis for ψ = 0, so that we have s(0) = −p/2. Hence, the input-output function s is
s(ψ) = p
2π ψ −p
2
(1)
The expressions for the ﬁrst and second derivatives of s(ψ) with respect to ψ will be needed:
s′(ψ) = p/(2π)
and
s′′(ψ) = 0
(2)
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
3
PSfrag replacements
f
p
e
p
s
d
x
y
P
C
µ
u
v
δ
b2
b3
a4
ψ
θ
O1
O2
Figure 3: Parameterization of Slide-o-Cam
PSfrag replacements
O1
p
x
y
u
v
x
s(0)
Figure 4: Home conﬁguration of the mechanism
The cam proﬁle is determined by the displacement of the contact point C around the cam. The Cartesian
coordinates of this point in the u-v frame take the form [7]
uc(ψ) =
b2 cos ψ + (b3 −a4) cos(δ −ψ)
(3a)
vc(ψ) = −b2 sin ψ + (b3 −a4) sin(δ −ψ)
(3b)
with coeﬃcients b2, b3 and δ given by
b2 = −s′(ψ) sin α1
(4a)
b3 =
p
(e + s′(ψ) sin α1)2 + (s(ψ) sin α1)2
(4b)
δ = arctan
 −s(ψ) sin α1
e + s′(ψ) sin α1

(4c)
where α1 is the directed angle between the axis of the cam and the translating direction of the follower, positive
in the ccw direction. Considering the orientation adopted for the input angle ψ and for the output s, as depicted
in Fig. 3, we have
α1 = −π/2
(5)
We now introduce the nondimensional design parameter η, which will be extensively used:
η = e/p
(6)
Thus, from Eqs. (1), (2), (4a–c), (5) and (6), we ﬁnd the expressions for coeﬃcients b2, b3 and δ as
b2 =
p
2π
(7a)
b3 =
p
2π
p
(2πη −1)2 + (ψ −π)2
(7b)
δ = arctan
 ψ −π
2πη −1

(7c)
whence a ﬁrst constraint on η, η ̸= 1/(2π), is derived. An extended angle ∆is introduced [9], so that the cam
proﬁle closes. Angle ∆is obtained as a root of Eq. 3. In the case of Slide-o-Cam, ∆is negative, as shown in
Fig. 5. Consequently, the cam proﬁle closes within ∆≤ψ ≤2π −∆.
2.1
Pitch-Curve Determination
The pitch curve is the trajectory of the center O2 of the roller, distinct from the trajectory of the contact
point C, which produces the cam proﬁle. The Cartesian coordinates of point O2 in the x-y frame are (e, s), as
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
4
PSfrag replacements
O1
x
y
u
v
C
ψ
(a)
PSfrag replacements
O1
x
y
u
v
C
ψ
x, u
y, v
∆
-u
-v
(b)
PSfrag replacements
O1
x
y
u
v
C
ψ
x, u
y, v
∆
-u
-v
u
v
(c)
Figure 5: Orientations of the cam found when vc = 0: (a) ψ = ∆; (c) ψ = π; and (d) ψ = 2π −∆
depicted in Fig. 3. Hence, the Cartesian coordinates of the pitch-curve in the u-v frame are
up(ψ) =
e cos ψ + s(ψ) sin ψ
(8a)
vp(ψ) = −e sin ψ + s(ψ) cos ψ
(8b)
2.2
Geometric Constraints on the Mechanism
In order to lead to a feasible mechanism, the radius a4 of the roller must satisfy two conditions, as shown in
Fig. 6a:
• Two consecutive rollers on the same side of the roller-carrying slider must not interfere with each other.
Since p is the distance between the center of two consecutive rollers, we have the constraint 2a4 < p.
Hence the ﬁrst condition on a4:
a4/p < 1/2
(9)
• The radius b of the shaft on which the cams are mounted must be taken into consideration. Hence, we
have the constraint a4 + b ≤e, the second constraint on a4 in terms of the parameter η thus being
a4/p ≤η −b/p
(10)
Considering the initial conﬁguration of the roller, as depicted in Fig. 4, the v-component of the Cartesian
PSfrag replacements
O1
x
y
p
e
b
C
a4
(a)
PSfrag replacements
O1
x
y
p
e
b
b
C
a4
(b)
Figure 6: Constraints on the radius of the roller: (a) a4/p < 1/2; and (b) a4/p ≤η −b/p.
coordinate of the contact point C is negative in this conﬁguration, i.e., vc(0) ≤0. Moreover, from the expression
for vc(ψ) and for parameters b3 and δ given in Eqs. (3b), (7b & c), respectively, the above relation leads to the
condition:

p
2πa4
p
(2πη −1)2 + (−π)2 −1

sin

arctan

−π
2πη −1

≤0
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
5
Further, we deﬁne A and B as:
A =
p
2πa4
p
(2πη −1)2 + π2 −1 and B = sin

arctan

−π
2πη −1

Since (2πη −1)2 > 0, we have
p
(2πη −1)2 + π2 > π
(11)
Hence, A > p/(2a4) −1. Furthermore, from the constraint on a4, stated in Eq. (9), we have p/(2a4) −1 > 0,
whence A > 0. Consequently, the constraint vc(0) ≤0 leads to the constraint B ≤0. We rewrite the constraint
B, by using the trigonometric relation,
∀x ∈R, sin(arctan x) =
x
√
1 + x2
Hence the constraint vc(0) ≤0 becomes
B ≤0 ⇐⇒
−π
(2πη −1)
p
1 + π2/(2πη −1)2 ≤0
which holds only if 2πη −1 > 0. Finally, the constraint vc(0) ≤0 leads to a constraint on η, namely,
η > 1/(2π)
(12)
2.3
Pressure Angle
The pressure angle is deﬁned as the angle between the common normal at the cam-roller contact point C and
the velocity of the follower [10], as depicted in Fig. 3, where the presure angle is denoted by µ. The smaller
|µ|, the better the force transmission. In the case of high-speed operations, i.e., angular velocities of cams
exceeding 50 rpm, the recommended bounds of the pressure angle are within 30◦. Nevertheless, as it is not
always possible to have a pressure angle that remains below 30◦, we adopt the concept of service factor, which
is the percentage of the working cycle with a pressure angle within 30◦[9]. The service angle will be useful to
take into consideration these notions in the ensuing discussion, when optimizing the mechanism.
For the case at hand, the expression for the pressure angle µ is given in [10] as
µ = arctan
s′(ψ) −e
s(ψ)

Considering the expressions for s and s′, and using the parameter η given in Eqs. (1), (2a) and (6), respectively,
the expression for the pressure angle becomes
µ = arctan
1 −2πη
ψ −π

(13)
We are only interested in the value of the pressure angle with the cam driving the roller, which happens with
π ≤ψ ≤2π −∆
(14)
Indeed, if we start the motion in the home conﬁguration depicted in Fig. 4, with the cam rotating in the ccw
direction, the cam begins to drive the roller only when ψ = π; and the cam can drive the follower until contact
is lost, i.e., when ψ = 2π −∆, as shown in Figs. 5b & c.
Nevertheless, as shown in Fig. 7, the conjugate cam can also drive the follower when 0 ≤ψ ≤π −∆; there is
therefore a common interval, for π ≤ψ ≤π −∆, during which two cams can drive the follower. In this interval,
the conjugate cam can drive a roller with lower absolute values of the pressure angle. We assume that, when
the two cams can drive the rollers, the cam with the lower absolute value of pressure angle eﬀectively drives the
follower. Consequently, we are only interested in the value of the pressure angle in the interval,
π −∆≤ψ ≤2π −∆
(15)
We study here the inﬂuence of parameters η and a4 on the values of the pressure angle while the cam drives
the roller, i.e., with π −∆≤ψ ≤2π −∆, as explained above.
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
6
• Inﬂuence of parameter η: Figure 8 shows the inﬂuence of the parameter η on the pressure angle, with a4
and p being ﬁxed. From these plots we have one result: The lower η, the lower |µ|.
• Inﬂuence of the radius of the roller a4: a4 does not appear in the expression for the pressure angle, but
it inﬂuences the value of the extended angle ∆, and hence, the plot boundaries of the pressure angle, as
shown in Fig. 7.
By computing the value of the extended angle ∆for several values of a4, we can see that the higher a4, the
lower |∆|. Consequently, since the boundaries to plot the pressure angle are π −∆and 2π −∆, we can see
that when we increase a4, −∆decreases and the boundaries are translated toward the left, i.e., toward higher
absolute values of the pressure angle.
PSfrag replacements
µ
ψ
π
π −∆
2π −∆
-6
-2
2
6
10
100
50
-50
Figure 7: Pressure angle distribution
PSfrag replacements
µ
ψ
η = 5
η = 2
η = 1.5
η = 1
η = 0.8
η = 0.4
η = 1/π
8
6
4
2
-20
-40
-60
-80
Figure 8: Inﬂuence of parameter η on the pressure
angle µ (in degree), with p = 50 mm and a4 = 10 mm
3
Convexity of the Cam Proﬁle and Undercutting
In order to enhance machining accuracy, we need the cam proﬁle to be fully convex. In this section we establish
conditions on the design parameters η and a4 in order to have a fully convex cam proﬁle. We thus study the
sign-change in the curvature of the cam proﬁle via that of the pitch curve. Furthermore, for cam design in
roller-follower mechanisms, we should also consider undercutting. Undercutting occurs when the radius of the
roller is greater than or equal to the minimum absolute value of the radius of curvature of the pitch curve. Upon
avoiding undercutting, the sign of the curvature of the pitch curve is identical to that of the cam proﬁle.
3.1
Curvature of the Cam Proﬁle
The curvature of any planar parametric curve, in terms of the Cartesian coordinates u and v, and parameterized
with any parameter ψ, is given by [10]:
κ = v′(ψ)u′′(ψ) −u′(ψ)v′′(ψ)
[u′(ψ)2 + v′(ψ)2]3/2
(16)
The sign of κ in Eq. (16) tells whether the curve is convex or concave at a point: a positive κ implies convexity,
while a negative κ implies concavity at that point. To obtain the curvature of the cam proﬁle for a given roller-
follower, we use the Cartesian coordinates of the pitch curve, since obtaining its ﬁrst and second derivatives
leads to simpler expressions as compared with those associated with the cam proﬁle itself. Then, the curvature
of the cam proﬁle is derived by a simple geometric relationship between the curvatures of the pitch curve and
of the cam proﬁle.
The Cartesian coordinates of the pitch curve were recalled in Eqs. (8a & b), while Eqs. (2a & b) give their
ﬁrst and second derivatives as
s(ψ) = p/(2π)ψ −p/2 ,
s′(ψ) = p/(2π) ,
s′′(ψ) = 0
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
7
With the above-mentioned expressions, we can compute the ﬁrst and second derivatives of the Cartesian coor-
dinates of the pitch curve with respect to the angle of rotation of the cam, ψ:
u′
p(ψ)
=
[s′(ψ) −e] sin ψ + s(ψ) cos ψ
(17a)
v′
p(ψ)
=
[s′(ψ) −e] cos ψ −s(ψ) sin ψ
(17b)
u′′
p(ψ)
=
[2s′(ψ) −e] cosψ −s(ψ) sin ψ
(17c)
v′′
p(ψ)
=
−[2s′(ψ) −e] sin ψ −s(ψ) cos ψ
(17d)
By substituting η = e/p, along with Eqs. (17a–d), into Eq. (16), the curvature κp of the pitch curve is obtained
as
κp = 2π
p
[(ψ −π)2 + 2(2πη −1)(πη −1)]
[(ψ −π)2 + (2πη −1)2]3/2
(18)
which remains bounded by virtue of condition (12).
Let ρc and ρp be the radii of curvature of both the cam proﬁle and the pitch curve, respectively, and κc the
curvature of the cam proﬁle. Since the curvature is the reciprocal of the radius of curvature, we have ρc = 1/κc
and ρp = 1/κp. Furthermore, due to the deﬁnition of the pitch curve, it is apparent that
ρp = ρc + a4
(19)
Writing Eq. (19) in terms of κc and κp, we obtain the curvature of the cam proﬁle as
κc =
κp
1 −a4κp
(20)
with κp given in Eq. (18). As we saw previously, we want the cam proﬁle to be fully convex, which happens if
the pitch curve is fully convex too. We thus ﬁnd ﬁrst the convexity condition of the pitch curve.
3.2
Convexity Condition of the Pitch Curve
Considering the expression for κp in Eq. (18), we have, for every value of ψ, κp ≥0 if (2πη −1)(πη −1) ≥0
and η ̸= 1/(2π), whence the condition on η:
κp ≥0
if
η ∈[0, 1/(2π)[
∪
[1/π, +∞[
(21)
Figure 9 shows pitch-curve proﬁles and their curvatures for two values of η. The condition on η given in Eq. (21)
PSfrag replacements
-20
-10
-10
10
10
20
20
30
PSfrag replacements
-20
-20
-40
40
40
20
20
30
PSfrag replacements
ψ
κp
0.2
-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
-1.4
2
4
6
(a)
PSfrag replacements
ψ
κp
0.022
0.023
0.024
0.025
0.026
0
1
2
3
4
5
6
7
(b)
Figure 9: Pitch-curve proﬁles and their corresponding curvatures with p = 50 mm:
(a) η = 0.2 (η ∈]1/(2π), 1/π[) and (b) η = 0.7 (η > 1/π)
must be combined with the condition appearing in Eq. (12), η > 1/(2π); hence, the ﬁnal convexity condition
of the pitch curve is:
η ≥1/π
(22)
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
8
3.3
Undercutting-Avoidance
We assume in this subsection that the pitch curve is fully convex, i.e., κp ≥0 and η ≥1/π. In order to avoid
undercutting, i.e., in order to have both the cam proﬁle and the pitch curve fully convex, we need κc to be
positive. Considering the expression for the curvature of the cam proﬁle κc of Eq. (20), the condition to avoid
undercutting is 1 −a4κp > 0, whence the condition on the radius of the follower a4:
a4 <
1
κp(ψ)
∀ψ ∈R
Since κp is positive, this condition can be written as
a4 <
1
κpmax
,
with
κpmax = max
ψ∈R κp(ψ)
(23)
• Expression for κpmax: In order to compute the expression for κpmax, we need the ﬁrst derivative κ′
p of κp
with respect to ψ and its roots. With the condition η ≥1/π, the expression for κp given in Eq. (18) is
diﬀerentiable for every value of ψ. We thus obtain
κ′
p = −2π
p
(ψ −π)(ψ2 −2πψ + π2 + 4η2π2 −10ηπ + 4)
[(ψ −π)2 + (2πη −1)2]5/2
(24)
The roots of κ′
p are, apparently, ψ1 = π and the roots ψ2 and ψ3 of the polynomial
P(ψ) = ψ2 −2πψ + π2 + 4η2π2 −10ηπ + 4
Let βψ be the discriminant of the equation P = 0, i.e.,
βψ = −4η2π2 + 10ηπ −4
Therefore, the sign of βψ and, consequently, the roots ψ2 and ψ3, depend on the value of η. Let βη be the
discriminant of βψ = 0, a quadratic equation in η. Hence, βη = 9π2, which is positive. The two roots of βψ are
1/2π and 2/π. Thus,
βψ > 0
if
η ∈
 1
π , 2
π

or
βψ < 0
if
η ∈
 2
π , +∞

βψ = 0
if
η = 2
π
We now study the roots of κ′
p according to the value of η.
• η ∈[1/π, 2/π[: βψ > 0, and the polynomial P has two roots ψ2 and ψ3, so that κ′
p has three roots:
ψ1
=
π
(25a)
ψ2
=
π +
p
−4η2π2 + 10ηπ −4
(25b)
ψ3
=
π −
p
−4η2π2 + 10ηπ −4
(25c)
• η ∈]2/π, +∞[: βψ < 0, and the polynomial P has no real roots, so that κ′
p has only one root, ψ1 = π.
• η = 2/π: βψ = 0, and the polynomial P has one double root equal to π, so that κ′
p has one triple root
ψ1 = π.
To decide whether these roots correspond to minima or maxima, we need to know the sign of the second
derivative κ′′
p of κp with respect to ψ, for the corresponding values of ψ. If the second derivative is negative, the
stationary value is a maximum; if positive, a minimum. The expressions for the second derivatives, as derived
with computer algebra, for the values of ψ given in Eqs. (25a–c):
κ′′
p(ψ1) =
4π(ηπ −2)
p|2ηπ −1|3(2ηπ −1)
κ′′
p(ψ2) = κ′′
p(ψ3) =
8π(ηπ −2)
9p(2ηπ −1)√6ηπ −3
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
9
PSfrag replacements
κp
ψ (rad)
2
6
6
0.02
0.04
(a)
2
PSfrag replacements
κp
ψ (rad)
0.023
0.025
0.027
0
6
6
(b)
Figure 10: Pitch-curve curvature for p = 50 mm: (a) η = 1/π and (b) η = 2/π
• If η ∈[1/π, 2/π[, κ′′
p(ψ1) > 0 and κ′′
p(ψ2) = κ′′
p(ψ3) < 0, the curvature of the pitch curve has one local
minimum for ψ1 and two maxima, for ψ2 and ψ3. Hence, the value of κpmax is
κpmax1 = κp(ψ2) = κp(ψ3) =
4π
3p√6ηπ −3
(26)
Figure 10a shows a plot of the pitch-curve curvature with η = 1/π and p = 50 mm. Since η is taken equal
to the convexity limit 1/π, the curvature remains positive and only vanishes for ψ = π.
• If η ∈]2/π, +∞[, κ′′
p(ψ1) < 0, the curvature of the pitch curve has a maximum for ψ1. Hence, the value of
κpmax is
κpmax2 = κp(ψ1) = 4π
p
(2η2π2 −3ηπ + 1)
(4η2π2 −4ηπ + 1)3/2
(27)
Figure 9d shows a plot of the pitch-curve curvature with η = 0.7 (η > 2/π) and p = 50 mm.
• If η = 2/π, κ′′
p(ψ1) = 0, we cannot tell whether we are in the presence of a maximum or a minimum. We
solve this uncertainty graphically, by plotting the curvature of the pitch curve for η = 2/π and p = 50 mm.
Figure 10b reveals that the curvature has a maximum for ψ1. The value of this maximum can be obtained
by substituting η by 2/π into either κpmax1 or κpmax2, expressed in Eqs. (26) and (27), respectively.
In summary, to have a fully convex cam proﬁle, taking the geometric constraints on the mechanism into con-
sideration, parameter η must obey the condition given in Eq. (22), i.e. η ≥1/π. We combine the condition on
a4 to avoid undercutting, as given in Eq. (23), with the geometric constraints on the mechanism, as given in
Eqs. (9) and (10), which are, respectively, a4/p < 1/2 and a4/p ≤η −b/p:
if η ∈
 1
π , 2
π

,
a4
<
min

1
κpmax1
, p
2, ηp −b

(28a)
if η ∈
 2
π , +∞

,
a4
<
min

1
κpmax2
, p
2, ηp −b

(28b)
where κpmax1 and κpmax2 are given in Eqs. (26) and (27), respectively.
4
Optimization of the Roller Pin
We concluded in the previous section that the lowest values of parameters η and a4 led to the lowest values
of the pressure angle. Nevertheless, we must take into consideration that the smaller the radius of the roller,
the bigger the deformation of the roller pin, and hence, a decrease of the stiﬀness and the accuracy of the
mechanism. In this section we formulate and solve an optimization problem to ﬁnd the best compromise on
parameters η and a4 to obtain the lowest pressure angle values with an acceptable deformation of the roller pin.
In this section we adopt the approach proposed by Teng and Angeles [11].
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
10
4.1
Minimization of the Elastic Deformation on the Roller Pins
Here we ﬁnd the expression for the maximum elastic deformation on the pin, which will be minimized under
given constraints. Figure 11 displays the free part of the pin, i.e., the part not ﬁxed to the roller-carrying slider,
as a cantilever beam, where the load F =
q
f 2x + f 2y denotes the magnitude of the force f transmitted by the
cam. This force is applied at a single point at the end of the pin in the worst loading case. Although the
dimensions of the pin are not those of a simple beam, we assume below that the pin can be modeled as such,
in order to obtain an explicit formula for its deﬂection. This assumption was found to be plausible by testing
it with FEA [11].
PSfrag replacements
f
f
L
a5
x
y
y
z
Figure 11: Approximation of the roller pin as a cantilever beam
The displacement vL at the free end of the pin turns out to be
vL =
q
v2x + v2y = FL3
3EI
(29)
where E is the Young modulus and I = πa4
5/4, with a5 denoting the radius of the pin, is the polar moment of
inertia of the cross section. Moreover, vx and vy are the pin elastic displacements in the x- and y-directions,
respectively, at the free end.
Before proceeding, we prove that the vertical component fy of the transmitted force is constant, and hence,
we will consider only the magnitude of the x-component of vL. Since we assume that the mechanism undergoes
a pure-rolling motion, the force exerted by the cam onto the roller, denoted by f = [fx fy]T , passes through
the center of the roller, i.e., its line of action passes through points O2 and C, as depicted in Fig. 3. With a
constant torque τ provided by the motor, we have τ = ||f||d, where d denotes the distance from the center of
the input axis to the line of action of the force f. Moreover, we have d = b2 sin δ. Hence, τ = ||f||(b2 sin δ) =
b2(||f|| sin δ) = b2fy. Finally, since b2 = 2π/p we obtain the expression for fy sought:
fy = 2πτ
p
= F0
(30)
Since τ is constant, fy is also constant throughout one full turn of the cam. Consequently we only have to
consider the x-component of f, and hence, vx, as given below, should be minimized:
vx = |fx|L3
3EI
= 4|fx|L3
3Eπa4
5
= β |fx|
a4
5
,
β = 4L3
3Eπ
(31)
β thus being a constant factor. The objective function z, to be minimized, is thus deﬁned as
z = f 2
max
a4
5
→
min
η,a4,a5
(32)
where fmax is the maximum value of fx throughout one cycle. Since
fx =
fy
tan δ =
F0
tan δ
(33)
we obtain
z = 1
a4
5
max
ψ

f 2
x
	
= 1
a4
5
max
ψ
 F 2
0
tan2 δ

= F 2
0
a4
5
max
ψ

1
tan2 δ

D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
11
with δ, a function of ψ, given in Eq. (7c). Moreover, the system operates by means of two conjugate mechanisms,
which alternately take over the power transmission. We established in Eq. (15) that when one mechanism is in
positive action, ψ is bounded between ψi = π −∆and ψf = 2π −∆, which corresponds to δ bounded between
δi and δf with 0 ≤δi < δf ≤π. Moreover, functions 1/ tan2 δ and cos2 δ are both unimodal in −π ≤δ ≤0 and
in 0 ≤δ ≤π, their common maxima ﬁnding themselves at −π, 0 and π. Since cos2 δ is better behaved than
1/ tan2 δ, we redeﬁne z as
z = 1
a4
5
max
δi≤δ≤δf

cos2 δ
	
Furthermore, the function cos2 δ attains its global minimum of 0 in [0, π], its maximum in the interval [δi, δf],
included in [0, π], occurring at the larger of the two extremes of the interval, δi or δf. It follows that the
objective function to be minimized becomes
z = 1
a4
5
max

cos2 δi, cos2 δf
	
−→min
ν,a4,a5
with δi and δf the values of δ for ψi = π −∆and ψf = 2π −∆, respectively. Using the expression for δ given
in Eq. (7c) and the trigonometric identity
cos(arctan x) =
1
√
1 + x2
we obtain the expression for cos2 δ:
cos2 δ =
(2πη −1)2
(2πη −1)2 + (ψ −π)2
(34)
Hence,
cos2 δi
=
(2πη −1)2
(2πη −1)2 + (ψi −π)2
(35a)
cos2 δf
=
(2πη −1)2
(2πη −1)2 + (ψf −π)2
(35b)
Furthermore, since ψi = π −∆, ψf = 2π −∆and ∆< 0, we have ψf > ψi > π, ψf −π > ψi −π > 0 and,
consequently, from Eqs. (35a & b), cos2 δi > cos2 δf and the objective function to minimize becomes
z = cos2 δi
a4
5
→
min
η,a4,a5
(36)
with cos2 δi given in Eq. (35a) and ψi = π −∆.
4.2
Geometric Constraints
Two neighboring pins cannot be tangent to each other, as depicted in Fig. 12, and hence the radius a5 of the
pin is bounded as
a5/p < 1/4
(37)
Furthermore, a4 and a5 are not independent. From the SKF catalogue, for example, we have information on
bearings available in terms of the outer radius D and the inner radius d, as shown in Fig. 13. We divide these
bearings into ﬁve series, from 1 to 5. Hence, each series can be represented by a continuous function. We chose
series 2, in which the basic dynamic load rating C lies between 844 and 7020 N. Furthermore, for series 2, the
relation between D and d can be approximated by a linear function D vs. d, D ≃1.6d + 10 (in mm). Since
D = 2a4 and d = 2a5, the above equation leads to
a4 ≃1.6a5 + 5
in mm
(38)
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
12
PSfrag replacements
p
p/2
a4 a5
Figure 12: Geometric constraint on the
roller-pin radius
10
10
20
20
30
40
50
60
70
0
15
5
Series 1
Series 2
Series 3
Series 4
Series 5
d (mm)
D (mm)
Figure 13: Dimensions of the SKF bearings
We deﬁne now two non-dimensional parameters α4 and α5:
α4 = a4/p
α5 = a5/p
(39)
where α5 can be derived from Eq. (38) as
α5 = (5/8)α4 −25/(8p)
with p in mm.
(40)
Moreover, we recall the geometric constraint deﬁned in Eqs. (22) and (28) which can be rewritten as
g1 = 1
π −η ≤0
(41a)
g2 = α4 −1
2 < 0
(41b)
g3 = α4 −
1
pκpmax
< 0
(41c)
g4 = α4 −η + b
p ≤0
(41d)
g5 = α5 −1
4 < 0
(41e)
with cos2 δi and α5 given in Eqs. (35a) and (40), respectively.
4.3
Results of the Optimization Problem
We solve the foregoing optimization problem using the fmincon function of Matlab which is base on semi
quadratic programming, using p = 50 mm and b = 9.5 mm. One solution is found, corresponding to η = 0.69
and a4 = 24.9992 mm, with a value of z = 249. The algorithm ﬁnds the values of η and a4 as big as possible
considering the constraints. For this solution, constraints (41d & e) are active. Nevertheless, as we saw in
Subsection 3.3. about the inﬂuence of parameters h and a4, we want these parameters to be as small as possible
in order to have low pressure angle values. For the solution found above, the service factor is equal to 0%.
Consequently, we must ﬁnd a compromise between the pressure angle and the roller-pin elastic deformation.
Table 1 shows solutions found by the optimization algorithm upon reducing the boundaries of η. Each time the
algorithm ﬁnds the corresponding value of a4 as big as possible, constraint (41d) becomes active. Recorded in
this table is also the corresponding maximum elastic deformation of the roller pin vLmax (its expression is derived
below), the minimum and the maximum absolute values of the pressure angle, |µmin| and |µmax|, respectively,
and the service factor, as deﬁned in Section 2.4. From Eqs. (29) and (31) we have
vL = β
a4
5
q
f 2x + f 2y
Using Eqs. (30) and (33), the above equation leads to
vL = βF0
a4
5
r
1 +
1
tan2 δ
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
13
which can be simpliﬁed by means of the expression for cos2 δi given in Eq. (35a) as
vLmax = βF0
a4
5
p
(2πη −1)2 + (ψi −π)2
|ψi −π|
(42)
with F0, β and a5 given in Eqs. (30), (31) and (38), respectively, and ψi = π−∆. In Table 1 we record the value
of vLmax with L = 10 mm, τ = 1.2 Nm (according to the Orthoglide speciﬁcations recalled in Section 1) and
E = 2×105 MPa. We conclude from Table 1 that for this cam proﬁle we cannot ﬁnd an acceptable compromise
between a low deformation of the roller pin, and hence a high stiﬀness and accuracy of the mechanism, and low
pressure angle values. Indeed, for an acceptable deformation of the roller pin, vLmax = 8.87 µm, obtained with
η = 0.38, the service factor equals 54.68%, which is too low. On the other hand, for an acceptable service factor
of 79.43%, obtained with η = 1/π ≈0.3183, the deformation of the roller pin is equal to 710.19 µm, which is
too high.
Figure 14 depicts a rapid protote of the optimum mechanism whith an intermediate value, namely, η = 0.35.
The motor and the two cams translate, while the follower and the rollers stay ﬁxed on the base.
η
a4 (mm)
a5 (mm)
z
vLmax (µm)
|µmin| (◦)
|µmax| (◦)
service factor (%)
0.69
24.99
12.50
249
0.09
42.11
80.68
0
0.5
15.5
6.56
2968
0.50
28.59
69.81
6.85
0.4
10.5
3.44
32183
4.32
20.31
57.99
46.68
0.39
10
3.12
45490
6.07
19.46
56.42
50.68
0.38
9.5
2.81
66659
8.87
18.61
54.78
54.68
0.37
9
2.50
102171
13.63
17.75
53.04
58.69
0.36
8.5
2.19
165896
22.31
16.89
51.22
62.69
0.35
8
1.87
290765
39.71
16.03
49.31
66.70
0.34
7.5
1.56
566521
79.18
15.17
47.31
70.72
0.33
7
1.25
1.29 106
186.06
14.31
45.21
74.73
1/π
6.41
0.88
4.68 106
710.19
13.31
42.64
79.43
Table 1: Results of the optimization problem, with p = 50 mm, b = 9.5 mm, L = 10 mm, τ = 1.2 Nm and
E = 2 × 105 MPa
Figure 14: Layout of the two coaxial conjugate-cam
mechanism on the same shaft
Figure 15: Layout of the three conjugate-cam mech-
anism on the same shaft
5
Novel Conjugate-Cam Mechanism
This Section describes a novel mechanism and its prototype, based on Slide-o-Cam, that enables us to decrease
considerably the pressure angle while meeting the Orthoglide speciﬁcations.
This mechanism is composed of three conjugate cams mounted either (i) on three parallel shafts (with
the rollers placed on one single side of the roller-carrier) or (ii) on a single shaft (with the rollers placed on
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
14
one single side of three parallel roller-carrier), as depicted in Fig. 15. Mounted on two guideways, one motor
provides the torque to one camshaft. In case (i), the torque is transmitted to the two other camshafts through a
parallelogram mechanism coupling them, whose detailed design is reported in [12]. Shown in Fig. 16 is a layout
of this case, with 1, 2 and 3 denoting the three cams.
PSfrag replacements
p
p/2
1
2
3
O1
x, u
y, v
u
u
v
v
s(2π/3)
s(4π/3)
y12
y13
120◦
120◦
Figure 16: Layout of the non-coaxial conjugate-cam mechanism on three parallel shafts
The proﬁle of each cam is as described in Section 2. The cams are mounted in such a way that the angle
between the u-axis of cams 1 and 2 is 120◦, while the angle between the u-axis of cams 1 and 3 is 240◦. According
to the conﬁguration of the mechanism depicted in Fig. 16, and denoting by y12 and y13 the distance between
the origin of 1 and 2, and between the origin of 1 and 3, respectively, we have
y12 = p/2 + p + s(2π/3) ,
y13 = p/2 + 2p + s(4π/3)
Using the expression of the input-output function s given in Eq. (1), we obtain
y12 = 4p/3 ,
y13 = 8p/3
(43)
Figure 17 shows the pressure angle variation for each cam vs. ψ, where 1, 2 and 3 denote the plot of the
pressure angle for cams 1, 2 and 3, respectively. Moreover, cams 2 and 3 are rotated by angles 2π/3 and 4π/3,
respectively, from cam 1. We can also consider that the plot for cam 3 that drives the follower before cam 1
in a previous cycle, referred to as 3’ in Fig. 17, is out of phase −2π/3 with respect to cam 1. As we saw in
Eq. (14), cam 1 can drive the follower within π ≤ψ ≤2π −∆, which corresponds in Fig. 17 to the part of the
plot 1 between points B and D.
Consequently, cam 2 can drive the follower within
π + 2π/3 ≤
ψ
≤2π −∆+ 2π/3
i.e.
5π/3 ≤ψ ≤8π/3 −∆
and cam 3 within
π + 4π/3 ≤
ψ
≤2π −∆+ 4π/3
i.e.
7π/3 ≤ψ ≤10π/3 −∆
which is equivalent to saying that cam 3 can drive the follower in a previous cycle, within
π −2π/3 ≤
ψ
≤2π −∆−2π/3
i.e.
π/3 ≤ψ ≤4π/3 −∆
The above interval corresponds in Fig. 17 to the part of the plot 3’ between points A and C. Consequently,
there is a common part for cams 1 (plot 1) and 3 (plot 3’) during which these two cams drive the follower,
namely, between points B and C, which corresponds to
π ≤ψ ≤4π/3 −∆
(44)
Moreover, during this common part, cam 3 has lower absolute pressure angle values than cam 1, and
hence, we consider that only cam 3 drives the follower. Consequently, cam 1 drives the follower only within
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
15
η
a4 (mm)
a5 (mm)
vLmax (µm)
|µmin| (◦)
|µmax| (◦)
service factor (%)
0.5
15.5
6.56
0.26
28.59
49.41
10.49
0.4
10.5
3.44
2.88
20.31
37.20
70.02
0.39
10
3.12
4.14
19.46
35.81
76.02
0.38
9.5
2.81
6.20
18.61
34.39
82.02
0.37
9
2.50
9.76
17.75
32.95
88.03
0.36
8.5
2.19
16.39
16.89
31.48
94.04
0.35
8
1.87
29.89
16.03
29.98
100
0.34
7.5
1.56
61.07
15.17
28.47
100
0.33
7
1.25
147.02
14.31
26.93
100
1/π
6.41
0.88
576.95
13.31
25.12
100
Table 2: Design parameters, roller pin deformation and pressure angle for the novel conjugate-cam mechanism
of Fig. 15, with p = 50 mm, b = 9.5 mm, L = 10 mm, τ = 1.2 Nm and E = 2 × 105 MPa.
4π/3 −∆≤ψ ≤2π −∆. These boundaries allow us to have a pressure angle lower than with coaxial conjugate
cams, since we do not use anymore the part of the cam proﬁle within π −∆≤ψ ≤4π/3 −∆, which has high
absolute pressure angle values. We can thus obtain a higher service factor for the mechanism.
The maximum roller-pin deformation vLmax derived in eq.(42) is reasonably low. In Table 2, we record
the values of η, a4, a5, vLmax, |µmin|, |µmax| and the service factor for the three-conjugate-cam mechanism of
Fig. 15. The best compromise is to use the three-conjugate-cam mechanism with η = 0.37, whence the radius of
the roller is a4 = 9 mm and the roller-pin deformation is vLmax = 9.76 µm with a good service factor of 88.03%.
6
Conclusions
The optimum design of Slide-o-Cam was made for two types of mechanism, i.e. with two and three-conjugate-
cams.
The three-conjugate-cam mechanism reported here allows us to drive the Orthoglide with prismatic
actuators using rotary DC motors. Moreover, the roller-pin deformation is minimized, and the service factor is
substantially increased with respect to the layout of Fig. 14. Further research is currently underway to evaluate
the inﬂuence of the number of lobes on each cam to increase the service factor using only two cams.
References
[1] Gonz´alez-Palacios, M. A. and Angeles, J., 2003, The design of a novel pure-rolling transmission to convert
rotational into translational motion”, ASME Journal of Mechanical Design, Vol. 125, pp. 205-207
o
o
o
o
PSfrag replacements
µ
3’
1
2
3
20
40
60
80
0
-20
-40
-60
-80
A
B
C
D
-2
-4
2
4
6
8
10
10
Figure 17: Pressure angle for the three cams
D. Chablat and J. Angeles, submitted to Journal of Mechanical Design
16
[2] Chablat, D. and Wenger Ph. 2003, “Architecture Optimization of a 3-DOF Parallel Mechanism for Ma-
chining Applications, the Orthoglide,” IEEE Transactions on Robotics and Automation, Vol. 19/3, pp.
403-410, June.
[3] Lampinen, J., 2003, “Cam shape optimisation by genetic algorithm,” Computer-Aided Design, Vol. 35,
pp. 727-737.
[4] Zhang, Y. and Shin, J.-H., 2004, “A Computational Approach to Proﬁle Generation of Planar Cam Mech-
anisms,” ASME Journal of Mechanical Design, Vol. 126-1, pp. 183–188.
[5] Bouzakis, K.D., Mitsi, S. and Tsiaﬁs, J., 1997, “Computer-Aided Optimum Design and NC Milling of
Planar Cam Mechanims,” Int. J. Math. Tools Manufact., Vol. 37, No. 8, pp. 1131-1142.
[6] Carra, S., Garziera, R. and Pellegrini, M., 2004, “Synthesis of cams with negative radius follower and
evaluation of the pressure angles,” Mechanism and Machine Theory, Vol. 34, pp. 1017–1032.
[7] Gonz´alez-Palacios, M. A. and Angeles, J., 1993, Cam Synthesis, Kluwer Academic Publishers B.V., Dor-
drecht.
[8] Waldron, K. J. and Kinzel, G. L., 1999, Kinematics, Dynamics, and Design of Machinery, John Wiley &
Sons, Inc., New York.
[9] Lee, M.K., 2001, Design for Manufacturability of Speed-Reduction Cam Mechanisms, M.Eng. Thesis, Dept.
of Mechanical Engineering, McGill University, Montreal.
[10] Angeles, J. and L´opez-Caj´un,C., 1991, Optimization of Cam Mechanisms, Kluwer Academic Publishers
B.V., Dordrecht.
[11] Teng, C.P. and Angeles, J., 2004, “An optimality criterion for the structural optimization of machine
element,” ASME Journal of Mechanical Design, Accepted for publication.
[12] Renotte, J. and Angeles, J. 2003, “The Design of a Pure-Rolling Cam Mechanism to Convert Rotationnal
into Translational Motion,” Department of Mechanical Engineering and Centre for Intelligent Machines
Report, TR-CIM-03-06, August, McGil University, Montreal.
y
x
O1
C
u
v
-15
-10
-5
5
10
15
-10
10
20