Andrzej Koszewnik, Kamil Troc, Maciej Słowik DOI 10.

2478/ama-2014-0039
PID Controllers Design Applied to Positioning of Ball on the Stewart Platform

PID CONTROLLERS DESIGN APPLIED TO POSITIONING OF BALL

ON THE STEWART PLATFORM

Andrzej KOSZEWNIK*, Kamil TROC*, Maciej SŁOWIK*

*Faculty of Mechanical Engineering, Bialystok University of Technology, ul. Wiejska 45C, 15-351 Białystok, Poland

[email protected], [email protected], [email protected]

Abstract: The paper presents the design and practical implementation of PID controllers for a Stewart platform. The platform uses a re-
sistance touch panel as a sensor and servo motors as actuators. The complete control system stabilizing the ball on the platform is real-
ized with the Arduino microcontroller and the Matlab/Simulink software. Two processes required to acquire measurement signals
from the touch panel in two perpendicular directions 𝑿 and 𝒀, are discussed. The first process includes the calibration of the touch panel,
and the second process - the filtering of measurement signals with the low pass Butterworth filter. The obtained signals are used to design
the algorithm of the ball stabilization by decoupling the global system into two local subsystems. The algorithm is implemented in a soft real
time system. The parameters of both PID controllers (PID𝒙 and PID𝒚) are tuned by the trial-error method and implemented in the micro-
controller. Finally, the complete control system is tested at the laboratory stand.

Key words: PID Controller, Low Pass Butterworth Filter, Stewart-Gough Platform, Arduino Microcontroller, USB Port, Touch Panel Sensor

1. INTRODUCTION The most popular type of actuators used in parallel link ma-
nipulators are pneumatic or hydraulic actuators. Both types
of actuators have many advantages and their application in the
Parallel link manipulators known as the “Stewart-Gough” plat- system depends on the payload to weight ratio and on the dimen-
forms (see Fig.1) were investigated thoroughly in the last years. sions of the system. Since pneumatic devices are less sensitive to
Since their potential advantages as: high rigidity, high accuracy, temperature changes and contamination, a lot of researchers
and high load-carrying capacity they seem to be better in compar-
have tried controlling air cylinders using various methods.
ison to the conventional serial link manipulators. That is why many One of the most popular control methods is the PID technique.
researches often choose such structures as control plants. The Wang and Van Varseveld and Bone (Wang et al., 1999; Varseveld
first parallel mechanism was introduced by Gough and Whitehall
and Bone, 1997) used PID controllers with nonlinear compensa-
in 1956 year (Gough and Whitehall, 2012) as a tyre-testing ma- tion for servo-pneumatic actuators to control the position of the
chine. Next, the design of the platform was improved by Stewart Stewart platform. Weng and Xu (2013) also used the PID control-
in 1965. Then, the modernized structure was used as an aircraft ler with the LQG filter to control the position and to damp the
or vehicle simulation device. Parallel manipulators were used also vibration of the platform. Other researchers used non-traditional
as platforms for communication satellite dishes and telescopes
control methods. Urenizezius et al. applied the LQR technique.
positioning. Paul et al implemented the sliding mode control (Urniezius
and Geguzis, 2014; Paul et al., 1994; Shunmugham and Hayaka-
wa, 1997). Wang et al. (2009) used the optimal control method
based on the generalized natural frequencies of the system
to stabilize the platform.
The Stewart platform discussed in the current article is a small
size structure (the dimension of the fixed platform is 30 cm,
and of the moving platform – 11 cm) with the low payload
to weight ratio. The control system for a ball on the platform
is realized by six electrical servo motors, the resistance touch
panel and the PID controller. The choice of the PID control algo-
rithm has been dictated mainly by practice problems of controlling
the platform rather than by theoretical considerations of designing
the control law. Other control methods would be difficult to imple-
ment because the mathematical model of the platform does not
exist yet. Moreover, the PID control technique is a first control law
investigated at this laboratory stand and it is realized together with
a student. The design and experimental verification of other con-
Fig. 1. Electrically actuated Stewart-Gough platform trol methods will be the subject of future investigations.

214
 acta mechanica et automatica, vol.8 no.4 (2014)

The paper presents consecutive steps of the design process The 6-DOF Stewart platform shown in Fig. 2 consists of:
and practical implementation of the chosen control algorithm for a base frame, a moving platform, six legs, and components of the
the platform and the Butterworth filters for the measurement sig- control system. As actuators the NewPower XL-9HM servo-
nals. motors have been chosen and located on the frame base.
As a sensor the RES121-4W touchpad has been chosen
(http://www.buddyrc.com/new-power-xl-9hm.html). The process
2. THE STEWART PLATFORM
of stabilization of the ball on the upper platform is realized
AS A PLANT OF THE CONTROL SYSTEM by the 8-bit Arduino Mega 2560 microcontroller operating
at 16 MHz frequency.

3. THE CONTROL ALGORITHM

The control algorithm of stabilization of the ball on the upper

platform is realized by the established communication link be-
tween the Arduino microcontroller and the Matlab/Simulink soft-
ware via the USB port. The standard transmission rate for this port
is up to 480Mb/s (115 kb/s in the current case), so this rate
is sufficient to control the system in the soft real time system
regime (http://stackoverflow.com/questions/17308956/differences-
between-hard-real-time-soft-real-time-and-firm-real-time). The dia-
Fig. 2. The laboratory stand with the touch panel and the microcontroller grams of the control system and algorithm are shown in Figs. 3
and 4.

Fig. 3. The control scheme of the ball position

According to Fig. 4, the error signals in two perpendicular di- The complete control system is based on the Matlab/Simulink
rections 𝑋 and 𝑌 are obtained based on two measured signals software and the Arduino microcontroller hardware. As a result
(posx and posy) and two setpoint values (𝑟𝑒𝑓𝑥 and 𝑟𝑒𝑓𝑦 ). Both the block diagram consists four main subystems are prepared
desired values: 𝑟𝑒𝑓𝑥 = 113 and 𝑟𝑒𝑓𝑦 = 3 correspond to a real and shown in Fig.4:
middle point of the touch panel in two directions 𝑋 and 𝑌, respec-  read from touchpanel – the signals from touchpanel is read by
tively. If the ball’s position on the platform is greater than 2 mm Arduino and filtered,
in each direction then the control signals 𝑢𝑥 and 𝑦 are calculated  simultaneous start – synchronous switch on both control
in Matlab/Simulink and configuration of servo motors is changed feedback loop,
(the signals are sent to Arduino). Otherwise the configuration  write to servos – the values control signals from Matlab
of the servo motors remains the same. are recalculated and pass to servos,
 PID control – the control signals 𝑢𝑥 and 𝑢𝑦 are calculated
with using Matlab software.

215
Andrzej Koszewnik, Kamil Troc, Maciej Słowik DOI 10.2478/ama-2014-0039
PID Controllers Design Applied to Positioning of Ball on the Stewart Platform

Read value from axis X

START 900
Simulation

800 on the top

in the middle
NO switch 'start' is YES 700 at the bottom
turned on?
x y
600

Resistance [  ]
read position x read position y
500
don't control,
servos in current 400
position

filter signal x filter signal y 300

200
posx posy

100
setpoint refx setpoint refy 0 10 20 30 40 50 60 70 80 90 100
ex=|refx-posx| ey=|refy-posy|
for axis X for axis Y Samples [n]
Fig. 5. The calibration of the touch panel in 𝑋 axis

Read value from axis Y

NO error ex in axis X is error ey in axis Y is 900
higher than 2 mm? YES higher than 2 mm?

NO YES 800
ex<=2 & ey<=2
on the left side
700
TRUE in the midlle
on the right side
calculate calculate 600
don't change

Resistance [  ]
the control signal the control signal
servos position
in axis X in axis Y
500
Write to servos in
the proper
400
configuration

300

Fig. 4. Scheme algorithm of stabilization the ball on the platform 200

(𝑟𝑒𝑓𝑥 = 113 mm, 𝑟𝑒𝑓𝑦 = 83 mm)

100
0 10 20 30 40 50 60 70 80 90 100
Samples [n]

4. CALIBRATION PROCESS OF THE TOUCH PANEL Fig. 6. The calibration of the touch panel in 𝑌 axis

Filtering signal x
The calibration process of the touch panel for different location 300
of the ball on the platform is carried out in the paper. In this order un-filtering ZOOM
position [mm]

200
the ball is located into two extreme position and one position into filtering

centre point on the sensor versus two perpendicular directions 100

𝑋 and 𝑌, respectively. In results values of resistance of electrodes
on the touch panel are measured and recorded in Fig. 5 0
0 1 2 3 4 5 6
and Fig. 6. Finally, obtained results are collected and transformed Filtering signal y
into displacement in Tab. 1. 150
position [mm]

ZOOM
100

Tab. 1. Results of calibration process of the touch panel 50

un-filtering
Direction Resistance 𝛺 Displacement mm filtering
0
𝑋 207-875 0-232 0 1 2 3 4 5 6
time [s]
𝑌 219-814 0-166 Fig. 7. The comparison of non-filtering and filtering signals

The comparison of both figures (Fig. 5 and Fig. 6) shown cor- Filtering signal x
200
rectly response action of the touch panel in both directions, but un-filtering
position [mm]

Time delay
obtained values of resistance in Y direction have firmly distribu- 150 filtering

tion. Thus, in order to design correctly control law these signals

100
should pass by low pass 2-order Butterworth filter in frequency
range up to 𝜔𝑐𝑢𝑡𝑜𝑓𝑓 = 35 rad/s, which transfer function as fol- 50
4 4.2 4.4 4.6 4.8 5
lows (http://www.ece.uah.edu/corses/ee426/Butterworth.pdf): Filtering signal y
0.1225 120
𝐻(𝑠) = (1) un-filtering
𝑠 2 +0.495𝑠+0.1225
position [mm]

100
filtering
Taking into account Fig. 7 and Fig. 8 we can see that the sig- 80

nals from the touch panel in perpendicular directions are filtered, 60

however application of this filter in the control system cause ap- 40

4 4.2 4.4 4.6 4.8 5
pear time delay on level 40 ms. In results we obtained possibility time [s]
of design the control system only in soft real-time regime system. Fig. 8. The comparison of non-filtering and filtering signals in zoom scale

216
 acta mechanica et automatica, vol.8 no.4 (2014)

5. CONTROL STRATEGY ing to Fig. 3 consists three inputs: 𝑟𝑒𝑓𝑥 = 113 (desired value),
measurement value of position of the ball and signal without error.
The last signal is responsible for simultaneously switching
The stage of design of stabilization the ball on the platform re-
the control system between two directions for two cases. Once
quire to consider location of all six servo motors on the base
the ball is in the center point of the touch panel with extreme
platform. As first, the opposite site servo motors are coupled
values less from +/- 2mm, but in the second case button START
versus their location on the fixed base. Next, taking into account
is switch off. If the ball is in the center point of the touch panel
one couple of servo motors, the control signals 𝑈 + and 𝑈 − are
then values of the error signals are checked. If -2< 𝑒𝑥 <2 for 𝑋
generated by controllers for each of them. The control signals
direction then the output signal from logic gate NAN𝐷𝑥 equal 0
expressed by Eq. 2 consists value of 90 that represents the initial
and output value with logic gate OR also is 0 (in case that
value of the control signal. Based on this assumption the control
NAN𝐷𝑦 also equal 0). In results the error signal is 0 and all servo
signals can be written as follows:
motors have still actual values. In second case the error signal
𝑈1+ (𝑡) = 90 + 𝑢𝑃𝐼𝐷𝑥 (𝑡) is omitted if mentioned button has been switched off. In such case
(2) variable posx equals -20 and in the result one of inputs of multiply
𝑈1− (𝑡) = 90 − 𝑢𝑃𝐼𝐷𝑥 (𝑡)
block is equal 0. Finally, the control signals 𝑋1 and 𝑋2 for 𝑋 axis
𝑑𝑒(𝑡) are generated by PID𝑥. Obtained signals in range of 30-150
where: 𝑢𝑃𝐼𝐷𝑥 (𝑡) = 𝑘𝑝 𝑒(𝑡) + 𝑘𝑖 ∫ 𝑒(𝑡)𝑑𝑡 + 𝑘𝑑 .
𝑑𝑡
corresponds to minimal and maximal pitch angle of rudder bar
Practical realization of control subsystem to stabilize position
of the servo motors.
of the ball in the 𝑋 axis is shown in Fig. 9. The subsystem accord-

Fig. 9. The control scheme of the control of ball position’s in X direction

The parameters of PID controllers are chosen directly on the 120

control signal- x and error- x

lab stand in few stages. In first order the system is prepared to

100
tests by setting up the rudder bar of actuators in parallel orienta-
error[mm], control signal [deg]

tion versus the fixed base of the platform. In result the initial condi- 80

tion of the control system and initial value of control signals 60

control signal- axis X
are set up to 0 and 90, respectively. In the next step the ball is 40 error- axis X
located in position (-55; -30) from middle position of the touch
20
panel. After that the parameters of the both controllers PID𝑥
and PID𝑦 are tuned by the trial error method. As a result 0

the transfer function of these controllers are expressed as follows: -20

0.1𝑠 2 +0.35𝑠+0.25 -40

𝑃𝐼𝐷𝑋 (𝑠) = (3)
𝑠 -60
0 1 2 3 4 5 6 7 8
0.12𝑠 2 +0.45𝑠+0.25 time[s]
𝑃𝐼𝐷𝑌 (𝑠) = (4)
𝑠
Fig. 10. The control signal and error signal derived by the PID𝑥 controller
Next, obtained signals from controllers are passed to micro-
controller Arduino. The results of carried out investigations – the
We can see, that the position of the ball is constant in desired
signals from both control subsystem PID𝑥 and PID𝑦 (error signal
point of the platform up to 2 second. After that, the both control
and control signal) were recorded and are shown in Fig. 10
loops are switched on as a result of change of the ball location
and Fig. 11, respectively.
greater than 2 mm in two directions. One more time stabilization
of the ball of the platform is achieved after 3 or 5 seconds.

217
Andrzej Koszewnik, Kamil Troc, Maciej Słowik DOI 10.2478/ama-2014-0039
PID Controllers Design Applied to Positioning of Ball on the Stewart Platform

The difference between both transient periods of stabilization REFERENCES

of the ball occurs because the controller PID𝑦 generates peaks
of high values after 4 seconds in the control signal. Finally, ob- 1. Gough V. E., Whitehall S. G. (2012), Universal tyre test machine,
tained results are satisfying in point of view of control system such Proceedings of the 9th International Technical Congress F.I.S.I.T.A,
structure and further investigations allow us to compare PID con- 117–137,1962.
troller with other types of control methods. 2. Paul A. K., Mishra J. K., Radke M. G. (1994, Reduced order sliding
mode control for pneumatic actuator, IEEE Transactions on Control
control signal- y and error- y
Systems Technology, Vol. 2, No. 3, 271–276.
140 3. Shunmugham P., Hayakawa Y. (1997), Practical design of adaptive
120
model-based sliding mode control of pneumatic actuators,
Proceedings of Advanced Intelligent Mechatronics, IEEE, Tokyo,
Japan, 140.
error[mm],control signal [deg]

100

80 4. Urniezius R., Geguzis E. (2014), Hybrid Fuzzy Logic and Adpative

LQR Controller for Swing-up, Positioting and Stabilization of Inverted
60
control signal- axis Y Pendulum, Elektronika ir Elektrotechnika, Vol.20, No.3, 11-15.
40 error- axis Y 5. Varseveld R. B. V., Bone G. M. (1997), Accurate position control
20
of a pneumatic actuator using on/off solenoid valves, IEEE/ASME
Transactions on Mechatronics, Vol. 2, No. 3, 195–204.
0 6. Wang J., Pu J., Moore P. (1999), A practical control strategyfor
-20 servo-pneumatic actuator systems, Control Engineering Practice,
Vol. 7, No. 12, 1483–1488.
-40
0 1 2 3 4 5 6 7 8 7. Wang W., Yang H., Zou J., Ruan X., Fu X. (2009) Optimal design
time[s] of Stewart platforms based on expanding the control bandwidth while
Fig. 11. The control signal and error signal derived by the PID𝑦 controller considering the hydraulic system design, Journal of Zhejiang
University, Vol.10(1), 22-30.
8. Weng Ch., Xu Z. (2013), Track-position and vibration control
6. CONCLUSIONS simulation for strut of the Stewart platform, Applied Physics and
Engineering, Vol.14(4), 281-291.
9. http://stackoverflow.com/questions/17308956/differences-between-
The research work aimed stabilization of the ball on the Stew- hard-real-time-soft-real-time-and-firm-real-time (10.09.2014)
art platform with using Arduino microcontroller and Matlab/Si- 10. http://www.buddyrc.com/new-power-xl-9hm.html (12.03.2014)
mulink software. The process of designation of the parameters 11. http://www.ece.uah.edu/corses/ee426/Butterworth.pdf (30.09.2014)
of the PID controllers is carried out in many stages. In the first
step we successfully shown two process of preparing of the
measurment signals: calibration of the touch sensor and filtering This work is co-financed from Dean’s Project W/WM/19/2013
signals with using low pass Butterworth filter. Next, the parame- and Statutory Work of Department of Automatic Control and Robotics
ters of both PID controllers (PID𝑥 and PID𝑦) are determined S/WM/1/2012
directly on the laboratory stand. Moreover, we passed these
parameters of the controllers to microcontroler Arduino and we
tested the whole control system in regime of soft-real time system.

218