Asymmetric S-Curve Generation and Tracking

Under Motion Constraints: An Experimental

2022 Advances in Science and Engineering Technology International Conferences (ASET) | 978-1-6654-1801-0/22/$31.00 ©2022 IEEE | DOI: 10.1109/ASET53988.2022.9735039

Validation
Elif Cicek Türker Türker Onur Akbati
ALTINAY Robot Technologies Inc. School of Engineering and Technology Dept. of Control and Automation Eng.
Istanbul Industry and Trade Free Zone Al Dar University College Yildiz Technical University
Istanbul, Turkey Dubai, UAE Istanbul, Turkey
[email protected] [email protected] [email protected]

Kamil Vedat Sancak

ALTINAY Robot Technologies Inc.
Istanbul Industry and Trade Free Zone
Istanbul, Turkey
[email protected]

Abstract—S-curve is one of the most applied approaches trajectories [3]. Similarly, a point-to-point trajectory generation
for trajectory generation for the point-to-point movement in based on an exponential function leading to obtaining a
industrial drives. In this note, the asymmetric S-curve structure is trapezoidal-like velocity profile and continuous jerk signal
addressed to develop a single-axis motion structure, and an algo-
rithm is introduced to generate the desired motion. Furthermore, is proposed in [4]. Furthermore, a level-shifted sinusoidal
a cascade P/PI position controller structure with a feed-forward waveform for the acceleration profile is offered to reduce the
path is suggested and the determination of the controller gains vibration and the settling time leading to an s-shaped trajectory
is discussed. Aiming to improve the smoothness of the desired with a smoother jerk profile [5]. The generalization of S-curve
trajectory and to reduce the tracking error, the proposed scheme trajectories in a recursive form in order to generate the motion
containing both trajectory generation and controller structure
is implemented experimentally in a ball screw linear motion profile in a systematic way is addressed in [2]. It is worth
system. The mechanical model parameters in that experimental mentioning that the jerk must take the maximum possible
setup, and accordingly the controller gains, are obtained based value for the time-optimal trajectory corresponding to S-curve
on an estimation algorithm. The acquired results show that and bang-bang control structure [6].
the proposed controller structure with the asymmetric S-curve The demanded acceleration and deceleration constraints
generation algorithm is convenient for practical applications.
Keywords—AS-curve, trajectory generation, tracking control may be different in some applications leading to asymmetric
S-curves (AS-curve). The generated motion profile is not sym-
metric with respect to time in such trajectories, and, in general,
I. I NTRODUCTION
this is required to obtain a trajectory that can be tracked with
Implementing robotics and automation systems in industrial lower vibration [7]. An AS-curve can be generated based
processes has escalated with the technological developments, on the S-curve structure by introducing an asymmetricity
and more advanced systems have been demanded the appli- parameter that is the ratio between the maximum values of
cations requiring high-performance positioning. Point-to-point the acceleration and deceleration [8]. In [9], a formulation is
trajectory planning is one of the topics studied in that field presented to generate AS-curve, and three different cases are
in order to generate sufficiently smooth trajectories, moving considered in terms of the amount of the distance to determine
the position from an initial value to a final value. Vibration the motion variables.
minimization, accuracy, optimization and smoothness are de- The complete set of equations for each segment of an AS-
manded significantly in such a motion under the existence of curve in terms of position, velocity, acceleration and jerk is
motion constraints [1], [2]. given in [12]. After formulation, the timing instances and the
S-curve is a well-known trajectory planning method utilized jerk value to be employed are determined based on the target
to generate S-formed point-to-point path. Generally, constant position. Similarly, the relevant calculations are performed for
jerk regions are employed in an S-curve produced for the constant jerk value and four different instances are defined for
position reference which can be differentiated twice contin- different forms of acceleration and velocity depending on their
uously. That type is called a cubic polynomial reference. In maximum values in [11], [13].
addition, there are higher-order position curves, trigonometric In this paper, a time-optimal straight-forward algorithm is
S-curve and sigmoid jerk models utilized to form more smooth developed to reach the desired position under the max/min

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.
 jerk, acceleration and velocity constraints based on the AS-
curve formulation proposed in previous studies. In addition, a
complete algorithm to determine the AS-curve parameters for Algorithm 1 AS-curve generation algorithm
all the different trajectory forms is presented that is convenient Require: j0 , Am , Vm , Pt and r
for practical applications. Furthermore, a feed-forward con- Tj ← Am /j0
troller structure is designed to track the desired time-varying Ta ← Vm /Am − Tj
position reference that is the result of the introduced AS-curve if Ta < 0 then ⊲ Am is not reachable
generation algorithm. Finally, various experiments for different Ta ← p0
trajectories are conducted and the results of those experiments Tj ← Vm /j0
show that the proposed algorithm works well and the target end if
performance is achieved. Tv ← j0 Tj (TPjt +Ta ) − r+12 ((2Tj + Ta ))
if Tv < 0 then ⊲ Vm is not reachable
II. AS- CURVE GENERATION ALGORITHM
Tv ← 0
A typical asymmetric S-curve trajectory is presented in λ ← Tj2 + j0 T8P t
j (r+1)
Fig. 1. Notice that the trajectory given for the position here is √
Ta ← λ − 3Tj
continuously differentiable three times which makes it suffi- if Ta < 0 then ⊲ Am is not reachable
ciently smooth for many applications. The curve is asymmetric Ta ←  0
since the maximum acceleration and deceleration values can Pt
 (1/3)
be assigned differently, and this is achieved by varying jerk Tj ← j0 (r+1)
values in different phases considering the constant maximum end if
jerk value allowed. This trajectory can be examined in seven end if
segments each defined for a constant value of the jerk. For return Tj , Ta , Tv
each segment, those values can be given by


 j0 , t 0 ≤ t < t1




 0 t 1 ≤ t < t2

−j0 , t 2 ≤ t < t3



J(t) = 0, t 3 ≤ t < t4 (1)
−j0 /r2 , t4 ≤ t < t5


j


 0



 0, t 5 ≤ t < t6
j0 /r2 ,

t 6 ≤ t < t7
J(t)
where t is the time, t0 is the initial time, t1 , . . . , t6 are the time
instants for segment changes, t7 is the final time, j0 fixed jerk
value for acceleration phase, r is the ratio between maximum Am
acceleration and maximum deceleration values. Am and Vm
denote the maximum values of acceleration and velocity,
respectively, and Pt is the target position. In addition, Tj , A(t)
Ta and Tv are defined as t1 − t0 = t3 − t2 , t2 − t1 and
t4 − t3 , respectively. It is worth mentioning here that Ta or
Tv , or both of them can be zero depending on the constraints V
m
and the target position. Integrating (1), one can obtain the
change of acceleration (A), velocity (V ) and position (P ) with V(t)
respect to time assuming zero initial conditions for velocity
P
and acceleration (see i.e. [12]). As a result of that, the final t
position can be determined as
j0 Tj (Tj + Ta ) P(t)
Pt = P (t0 ) + [(r + 1)(2Tj + Ta ) + 2Tv ] .
2
(2) Tj Ta Tj Tv rTj rTa rTj
Once the initial and final positions are given, this equation can
t t t t t t t t
be solved for Tj , Ta and Tv under the presence of the velocity 0 1 2 3 4 5 6 7
and acceleration constraints. In this manner, Algorithm 1 can
be given for a positive Pt assuming P (t0 ) = 0 to find Tj , Ta
and Tv . Notice that, this algorithm can be modified slightly Fig. 1. Typical asymmetric S-curve trajectory.
to obtain the time duration for each and every segment for a
given negative Pt .

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.
 Fig. 2. Control Scheme

mechanism can be given as

τ = Jα + Bc sgn(ω) + Bv ω (3)
where τ , J, Bc and Bv are the applied torque, moment of
inertia, coulomb and viscous friction coefficients, respectively.
And, α is the acceleration, ω is the velocity of the system,
sgn(.) is the standard signum function.
Considering physical limitations, the system was tested
Fig. 3. Identification Block Scheme using velocity closed-loop scheme, illustrated in Fig. 3. In
that manner, a proportional controller with Kpv = 0.0116,
and sinusoidal reference velocity signals with varying ampli-
III. C ONTROLLER S TRUCTURE tude between 40 − 135mm/s and frequency 1 − 4Hz were
applied. The inertia, coulomb friction and viscous friction
The plant characteristics and the designed control structure coefficients were estimated by using recursive least square
as well as the gain adjustment are introduced in this section. algorithm as J = 5.57 × 10−4 kgm2 , Bc = 0.12N m,
The control system of the ball-screw mechanism, given in Bv = 0.001N ms/rad, respectively.
Fig. 2, consists of a cascade P/PI position controller that After the estimation of system parameters, employing the
includes a proportional (Kpp ) position and a PI (Kpv , Kiv ) closed-loop transfer function of velocity control loop for
velocity feedback loop. The velocity and acceleration refer- damping ratio ζ = 0.5 and settling time Ts = 0.1, velocity PI
ences are also included via the feed-forward path in order to controller gains were obtained as Kpv = 0.0436 and Kiv =
improve the output performance. 3.5645. The velocity response resulted from the simulation
model with the given PI gains is given in Fig. 4. Low-pass
filter introduced in Fig. 2 was designed to cancel the zero of
Velocity Closed Loop Simulation Results
1.2 the closed-loop transfer function of the velocity loop and to
eliminate the additional overshoot. The position proportional
1 gain (Kpp ) was calculated from the overall closed-loop transfer
function. As per the Routh-Hurwitz stability criterion, the
0.8
system is stable as long as 0 < Kpp < (B + Kpv )/J, and
0.6 the gain was selected as Kpp = 40 accordingly.
Velocity

0.4 IV. E XPERIMENTAL R ESULTS

The designed controller and the AS-curve algorithm were
0.2
tested on a ball screw linear motion system. The system
0 consists of an Industrial PC (Beckhoff) and an AC servo motor
(Beckhoff, AM8033-1E21) with a servo driver (Beckhoff,
-0.2
0 2 4 6 8 10
Time (seconds)
TABLE I
BALL S CREW PARAMETERS
Fig. 4. Velocity response.
Diameter [mm] 25
Before the application of the introduced control scheme, Lead [mm] 5
Length [mm] 1100
a system identification procedure was performed to estimate Cart Travel Distance (approx) [mm] 800
the system parameters. Dynamic model of the ball screw

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.
 Desired Trajectory Exp #1
500

Jerk [mm/s 3]
0

-500
0 1 2 3 4 5 6 7

Acceleration [mm/s2]
200

0
Fig. 5. Experimental Setup
-200

0 1 2 3 4 5 6 7

AX8206). The specifications of ball-screw mechanism is given

Velocity [mm/s]
in Table I and the system structure is depicted in Fig. 5. In 200

order to verify the applicability of given AS-curve algorithm, 100

two different experiments were conducted for two different
cases. 0
0 1 2 3 4 5 6 7
Experiment #1: For the first experiment, the system was
800
started with the initial position Pi = 10 mm and jerk,

Position [mm]
600
maximum acceleration, maximum velocity and target position 400
values were taken as j0 = 500 mm/s3 , Am = 200 mm/s2 , 200
Vm = 180 mm/s and Pt = 780 mm, respectively, with the rate 0
0 1 2 3 4 5 6 7
value r = 1. Obtained desired jerk, acceleration, velocity and Time [s]
position trajectories are presented in Fig. 6. Note that, the AS-
curve in this experiment contains all segments with Tj = 0.4s,
Ta = 0.5s and Tv = 2.9778s corresponding to constant jerk, Fig. 6. Desired Trajectory for Experiment 1.
acceleration and velocity regions. It is also clear in Fig. 7 that
the given trajectory is followed with an acceptable position
and velocity tracking errors and the required control signal is
verified.
Experiment #2: Initial position of the system was selected
as Pi = 650 mm in the second experiment, and the trajectory 10 -3 Results Exp #1

was calculated for the jerk, maximum acceleration, maximum

Position Error [mm]

velocity and target position values of j0 = 100 mm/s3 , Am =

0
300 mm/s2 , Vm = 200 mm/s and Pt = 250 mm, respectively,
with the rate value, r = 2. -5

The calculated trajectory in this experiment is given in -10

Fig. 8, and it consists of four segments only where Ta = 0 s 0 1 2 3 4 5 6 7

and Tv = 0 s. The trajectory is asymmetric with the ac-

celeration time Tj = 1.1006 s, and the deceleration time
Velocity Error [mm/s]

0.5
rTj = 2.2012 s. Also, one can observe that the acceleration 0
and velocity values do not reach their maximum values.
-0.5
Position and velocity tracking errors are given in Fig. 9 with
-1
the applied control signal for this experiment.
Detailed position and velocity graphs are illustrated in 0 1 2 3 4 5 6 7

Fig. 10-13 for both of the experiments. It can be concluded

from these results that the system can follow the trajectory 0.4
Control Signal [Nm]

with minor tracking errors showing the achievement of the 0.3

proposed structure. It is worth to mention here that the actual 0.2

velocity of the system was obtained by differentiating the
0.1
position value numerically and the tracking performance can
0
be improved with the implementation of appropriate filter 0 1 2 3 4 5 6 7
structure to eliminate the effect of the high frequency noise. Time [s]

Finally, the maximum and RMS values of position and velocity

tracking errors calculated for t = 0.15−6.58 [s] in Experiment
#1 and t = 0.15 − 7.6 [s] in Experiment #2 are presented in Fig. 7. Results for Experiment 1.

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.
 Desired Position vs Actual Position Exp #1

Position [mm]
Desired Trajectory Exp #2 Desired Position
200 600 Actual Position
400
Jerk [mm/s 3]

200
0
0
0 1 2 3 4 5 6 7
Time [s]
-200
0 1 2 3 4 5 6 7 8 11.5 780.01

780
Acceleration [mm/s2]

100

Position [mm]
Position [mm]
11
779.99
0
779.98
10.5
-100 779.97

0 1 2 3 4 5 6 7 8 10 779.96
0 0.05 0.1 0.15 0.2 5.5 6 6.5
0 Time [s] Time [s]
Velocity [mm/s]

-50 Fig. 10. Desired & actual position in experiment 1.

-100

-150
0 1 2 3 4 5 6 7 8 Table II (the transient period is excluded).
Position [mm]

600 Desired Velocity vs Actual Velocity Exp #1

200

Velocity [mm/s]
Desired Velocity
Actual Velocity
400
100

200
0 1 2 3 4 5 6 7 8
0
Time [s] 0 1 2 3 4 5 6 7
Time [s]
10 181

Velocity [mm/s]
Velocity [mm/s]

180
Fig. 8. Desired Trajectory for Experiment 2.
6
179
4
178
2

0 177
0 0.05 0.1 0.15 0.2 1.2 1.4 1.6 1.8 2
Time [s] Time [s]

10 -3 Results Exp #2
10
Fig. 11. Desired & actual velocity in experiment 1.
Position Error [mm]

TABLE II
0 C ONTROL S YSTEM P ERFORMANCE

Exp #1 Exp #2
0 1 2 3 4 5 6 7 8
Position Error [mm] 0.0079 0.0047
MAX
Velocity Error [mm/s] 0.8529 0.8888
2
Position Error [mm] 0.002 0.0012
RMS
Velocity Error [mm/s]

1.5 Velocity Error [mm/s] 0.1993 0.1678

1

0.5

-0.5
V. C ONCLUSION
0 1 2 3 4 5 6 7 8 An AS-curve generation algorithm and a controller driving
the output to track the desired position reference for point to
0
point motion have been presented in this paper. The developed
Control Signal [Nm]

-0.05 structure has been applied to a ball-screw mechanism exper-

-0.1 imentally in order to test its achievement. Two experiments
-0.15 have been conducted for different trajectories to emphasize
-0.2 the constraints, and satisfactory results have been obtained.
-0.25
0 1 2 3 4 5 6 7 8
Experimental results have revealed the viability of the pro-
Time [s] posed trajectory generation and tracking system for industrial
applications. The application of the given AS-curve algorithm
to more complicated trajectories such as an arc or circle path
Fig. 9. Results for Experiment 2. involving more than one degree-of-freedom is scheduled as

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.
 Desired Position vs Actual Position Exp #2 [5] Li, H., Le, M. D., Gong, Z. M., Lin, W., ”Motion profile design to
Position [mm] 600 Desired Position reduce residual vibration of high-speed positioning stages”, IEEE/ASME
Actual Position
500
Transactions On Mechatronics, vol. 14, no. 2, pp. 264–269, 2009.
[6] Wang S. D. et al., ”Planning method for multi-axis point-to-point
400
synchronization based on time constraints”, IEEE Access, vol. 8, pp.
300 85575–85604, May 2020.
200 [7] Li, H., ”A jerk-constrained asymmetric motion profile for high-speed
0 1 2 3 4 5 6 7 8
Time [s]
motion stages to reduce residual vibration”, International Journal of
250.015 Computer Applications in Technology, vol. 53 no. 2, pp. 149–156, 2016.
649.95 [8] Rew, K. H., Ha, C. W., Kim, K. S., ”A practically efficient method
Position [mm]
Position [mm]

250.01
649.9
for motion control based on asymmetric velocity profile”, International
250.005 Journal of Machine Tools and Manufacture, vol. 49, no. 7-8, pp. 678–
649.85
682, 2009.
250
649.8 [9] Rew, K. H., Kim, K. S., ”A closed-form solution to asymmetric
649.75 249.995 motion profile allowing acceleration manipulation”, IEEE Transactions
0 0.1 0.2 6.5 7 7.5 on Industrial Electronics, vol. 57, no. 7, pp. 2499–2506, 2009.
Time [s] Time [s]
[10] Bai Y., Chen X., Sun H., Yang Z., ”Time-optimal freeform s-curve
profile under positioning error and robustness constraints”, IEEE/ASME
Fig. 12. Desired & actual position in experiment 2. Transactions on Mechatronics, vol. 23, no. 4, pp. 1993–2003, May 2018.
[11] Zou, F., Qu, D., Xu, F., ”Asymmetric s-curve trajectory planning
for robot point-to-point motion” In IEEE International Conference on
future work. Accordingly, more advanced control structures Robotics and Biomimetics, pp. 2172–2176, 2009.
[12] Bai, Y., Chen, X., Yang, Z., ”A generic method to generate AS-
are also planned to be designed in future studies, especially curve profile in commercial motion controller”, In International Design
to deal with the uncertainties and disturbances existing in the Engineering Technical Conferences and Computers and Information in
closed-loop system. Engineering Conference, 2017.
[13] Zou, F., Qu, D., Wang, J., Xu, F., ”Asymmetric trajectory planning
for vacuum robot motion”, In 2011 IEEE International Conference on
Desired Velocity vs Actual Velocity Exp #2 Robotics and Automation, pp. 1-4, 2011.
0
Velocity [mm/s]

Desired Velocity
Actual Velocity
-50

-100

0 1 2 3 4 5 6 7 8
Time [s]
-112
0
Velocity [mm/s]

Velocity [mm/s]

-114

-116
-5 -118

-120
-10 -122
0 0.1 0.2 0.3 0.4 2 2.2 2.4 2.6 2.8 3
Time [s] Time [s]

Fig. 13. Desired & actual velocity in experiment 2.

ACKNOWLEDGMENT
This project is supported by The Scientific and Technolog-
ical Research Council of Turkey (TUBITAK) 1511 Research
Technology Development and Innovation Projects in Priority
Areas G. P. (Technology Focused Industrial Movement Pro-
gram) under Grant No: 1190131 and carried out in ALTINAY
Robot Technologies Inc., Istanbul.

R EFERENCES
[1] Mu, H., Zhou, Y., Yan, S., Han, A., ”Third-order trajectory planning
for high accuracy point-to-point motion”, Frontiers of Electrical and
Electronic Engineering in China, vol. 4, no. 1, pp. 83–87, 2009.
[2] Nguyen, K. D., Ng, T. C., Chen, I. M., ”On algorithms for planning
s-curve motion profiles”, International Journal of Advanced Robotic
Systems, vol. 5, no. 1, pp. 99–106, 2008.
[3] Fang, Y. et al., ”Smooth and time-optimal s-curve trajectory planning
for automated robots and machines”, Mechanism and Machine Theory,
vol. 137, pp.127–153, July 2019.
[4] Rymansaib, Z., Iravani, P., Sahinkaya, M. N.,”Exponential trajectory
generation for point to point motions”, In IEEE/ASME International
Conference on Advanced Intelligent Mechatronics, pp. 906-911, 2013

Authorized licensed use limited to: ULAKBIM UASL - YILDIZ TEKNIK UNIVERSITESI. Downloaded on February 20,2024 at 10:44:45 UTC from IEEE Xplore. Restrictions apply.