Developing low-cost two wheels balancing scooter using proportional derivative controller

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 3, June 2022, pp. 2454~2464
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i3.pp2454-2464  2454
Journal homepage: http://ijece.iaescore.com
Developing low-cost two wheels balancing scooter using
proportional derivative controller
Petrus Sutyasadi1
, Manukid Parnichkun2
1
Mechatronics Department, Faculty of Vocational, Sanata Dharma University, Yogyakarta, Indonesia
2
Mechatronics Department, School of Engineering and Technology, Asian Institute of Technology, Pathum Thani, Thailand
Article Info ABSTRACT
Article history:
Received Apr 29, 2021
Revised Dec 29, 2021
Accepted Jan 10, 2022
This study suggests a strategy for creating a low-cost two-wheeled balancing
scooter with a simpler system that can work similarly to the commercial
ones on the market. A simple system will help people to understand how it
works and how to build it. The mechanical parts were made from simple
hollow bar iron. Wheelbarrow wheels were attached to two electric bicycle
motors, and the controller was using an 8-bit microcontroller running a
proportional derivative (PD) controller. PD controller only is not enough to
run the scooter with a passenger smoothly. Some strategies were added to
overcome some non-linear problems due to the use of low-cost components.
Finally, the system is successfully built and can be ride by a 65 kgs weight
of rider. The scooter can turn left or right, and even to make a 360-degree
spot-rotation.
Keywords:
Balancing scooter
Proportional derivative
controller
Segway
Two wheels This is an open access article under the CC BY-SA license.
Corresponding Author:
Petrus Sutyasadi
Mechatronics Department, Faculty of Vocational, Sanata Dharma University
Yogyakarta, Indonesia
Email: peter@usd.ac.id
1. INTRODUCTION
Two wheels balancing platform is a platform that shows the principles of a dynamic control system.
It is one of the implementations of an inverted pendulum [1]. This kind of balancing platform is applied in
many applications, including transportation systems in the form of a two-wheeled balancing scooter. The
scooter uses several sensors to form a closed-loop control to know its position and orientation [2]. A
gyroscope and accelerometer provide information to calculate the orientation of vertical direction [3]. If there
is an encoder, it will check the rotation direction of the motor. This type of platform has good
maneuverability and requires a tiny space to make a turn [4]. The Segway Company makes a commercial two
wheels balancing scooter called Segway personal transporter (Segway PT). This product is quite strong
because it uses a brushless motor with neodymium magnets and powered by a 72 volts battery. It uses a
digital signal processing board from Delphi electronics and a signal processor from Texas instruments. It uses
a redundant system to anticipated system failure and keeps the rider in safety. It has a backup battery in case
the main battery power drops. With the backup battery, the rider can bring the Segway to some charging
spots available. A different prototype, a mobile inverted pendulum but uses a similar technique to balance the
system, is called JOE [5]. JOE is a two-wheeled balancing platform without a passenger.
Some researchers studied this balancing platform as an inverted pendulum model [6]–[9]. Other
researchers investigated balancing platforms on a ball instead of on wheels [10], [11]. This balancing
platform is a nonlinear system. However, a linear controller such as proportional–integral–derivative (PID)
can balance the system [12]–[16]. To get a better response in controlling a nonlinear system using PID, some
researchers use hybrid PID [17]–[19]. To enhance the use of PID in controlling the wheeled balancing

Int J Elec & Comp Eng ISSN: 2088-8708 
Developing low-cost two wheels balancing scooter using proportional derivative … (Petrus Sutyasadi)
2455
platform, fuzzy control is employed together with the PID control [20]. A fuzzy controller is used to calculate
the PID gain on the fly [21]. Another approach uses only a fuzzy controller [22] and a double loop fuzzy
controller [23] to control a self-balancing robot. Another approach is using optimized PID for a balancing
system [24], optimal control [25], [26] and adaptive control [27]. A comparison of an optimized PID and an
optimized fuzzy show that optimized fuzzy performs faster and less jerky [28]. Backpropagation neural
network is also capable of navigational control of a two-wheeled balancing robot in an unknown terrain [29].
A two-wheeled balancing platform as a personal transporter should provide stability on each of its
maneuvers during the riding. The most critical consideration in designing this type of scooter is stability.
That is due to the simplification of its modules, including the controller. All of the papers listed above say
little on solving weaknesses like backlash, low power motor drivers, or too much noise while using low-cost
sensors. These issues emerge as we create a rideable balanced scooter using low-cost parts. The remainder of
the paper is organized as follows; section 2 describes the research methodology. Section 3 shows the result
and analysis. Section 4 presents the conclusions.
2. RESEARCH METHOD
2.1. Mechanical design
The structure of the scooter is made from square hollow bar iron. The overall mechanical design is
made using SolidWork software. The construction consists of two major parts: the handlebar part and the
base part. Both of them are combined using bolt and nut at the lower part of the handlebar. Its controller,
direct current (DC) drivers, DC motors, sensors, and batteries are put on the base part. The base part is
connected to the wheels. The handlebar part is used to mount battery indicator, switches, and steering
potentiometer. There is an emergency killing switch as well. The switch is used to cut the power immediately
in case the system is not stable.
The center of gravity (CoG) of the system should lie on the center. All of the components, especially
those that have more weight should be put evenly. The controller can balance the system dynamically. But if
the CoG is lied rather to the front, the scooter will balance itself with the handlebar tilted back to maintain the
CoG position on the center. This situation is not comfortable at all for the passenger. Tilting the handlebar
forward or backward will move the system forward or backward. So, to ease the works of the controller, the
handlebar should be designed to stay vertical during the static balance. The scooter tends to move in the
direction where the system is statically unbalanced. The motor has a built-in gearbox with an 88:9 ratio.
Another ratio was added using a sprocket system with a 2:1 ratio. Figure 1 shows the mechanical design of
the scooter. Figure 2 shows the free body diagram of a two wheels balancing platform. The free body
diagram of the left and the right wheels are shown in Figure 2(a). While Figure 2(b) shows the free body
diagram of the chassis and the handle bar. The linearized state space equation is [30]:
[
𝑥̇
𝑥̈
𝜙̇
𝜙̈
] =
[
0 1 0 0
0
2𝑘𝑚𝑘𝑒(𝑀𝑝𝑙𝑟−𝐼𝑝−𝑀𝑝𝑙2)
𝑅𝑟2𝛼
𝑀𝑝
2𝑔𝑙2
𝛼
0
0 0 0 1
0
2𝑘𝑚𝑘𝑒(𝑟𝛽−𝑀𝑝𝑙)
𝑅𝑟2𝛼
𝑀𝑝𝑔𝑙𝛽
𝛼
0]
[
𝑥
𝑥̇
𝜙
𝜙̇
] +
[
0
2𝑘𝑚(𝐼𝑝+𝑀𝑝𝐼2−𝑀𝑝𝐼𝑟)
𝑅𝑟𝛼
0
2𝑘𝑚(𝑀𝑝𝐼−𝑟𝛽)
𝑅𝑟𝛼 ]
𝑉
𝑎 (1)
with
𝛽 = (2𝑀𝑤 +
2𝐼𝑤
𝑟2 + 𝑀𝑝)….𝛼 = [𝐼𝑝𝛽 + 2𝑀𝑝𝐼2
(𝑀𝑤 +
𝐼𝑤
𝑟
)]
Variables used on the (1) are listed: 𝑘𝑚 is torque constant, 𝑘𝑒 is back EMF constant, 𝑅 is armature resistant,
𝑉
𝑎 is applied voltage, 𝐼𝑤𝐼𝑤 is inertia of the wheels, 𝑟 is radius of the wheels, 𝐼𝑝𝐼𝑝 is inertia of the chassis, 𝜙 is
scooter’s leaning angle, 𝜙̇ is angular velocity of the scooter’s leaning angle, 𝜙̈ is angular acceleration of the
scooter's leaning angle, 𝑙 is distance between wheel and center of gravity, 𝑥 is horizontal displacement, 𝑥̇ is
horizontal velocity of the scooter, 𝑥̈ is horizontal acceleration of the scooter, 𝑀𝑤 is mass of the wheel, 𝑀𝑝 is
mass of the chassis, and 𝑔 is gravity. After inputting the system parameters into (1), the state space equation
and the transfer function are obtained as,
[
𝑥̇
𝑥̈
𝜙̇
𝜙̈
] = [
0 1 0 0
0 −0.06626 21.02 0
0 0 0 1
0 −0.03092 15.42 0
] [
𝑥
𝑥̇
𝜙
𝜙̇
] + [
0
02721
0
0.127
] 𝑉
𝑎 (2)

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 2454-2464
2456
𝐺(𝑠) = 𝐺1(𝑠) + 𝐺2(𝑠) (3)
𝐺1(𝑠) =
𝑋(𝑠)
𝑉𝑎
=
0.2721𝑠2+2.417𝑒−16𝑠−1.526
𝑠4+0.06626𝑠3−15.42𝑠2−0.3714𝑠
(4)
𝐺2(𝑠) =
𝜑(𝑠)
𝑉𝑎
=
0.127𝑠
𝑠3+0.06626𝑠2−15.42𝑠−0.3714
(5)
A PD controller applicable on a single-input single-output (SISO) system only. Therefore, only
scooter inclination transfer function will be considered. Figure 3(a) shows the overall control system block
diagram. Since the reference angle is zero, the block diagram can be arranged like in Figure 3(b) to make it
easier to analyze. Therefore, the equation become as in (6).
𝛷(𝑠)
𝐹(𝑠)
=
𝑃(𝑠)
1+𝐶(𝑠)𝑃(𝑠)
(6)
Figure 1. Mechanical design of the scooter
(a) (b)
Figure 2. Free body diagram of the balancing scooter (a) free body diagram of the wheels and (b) free body
diagram of the chassis and the handle bar
(a) (b)
Figure 3. Re-arrange of the scooter closed loop system block diagram to get inclination angle response from
an impulse disturbance (a) the closed loop block diagram of the system and (b) rearranged block diagram to
make it easier to be analyzed

2457
2.2. Electronic design
The electronic part of the system consists of a controller, DC motor driver, sensor, battery, user
interface, and battery indicator. Figure 4 shows the block diagram of the control system. The sensor signals
consist of angle and angle rate of the system refer to the vertical direction. A proportional derivative (PD)
algorithm was employed to calculate the output signal. The output signal was translated into pulse width
modulation (PWM) signal to rotate the DC motors.
Figure 4. Closed loop diagram block of the scooter
The tilt angle was provided by an accelerometer combined with a gyroscope packed into a single
vertical gyro VG400. The accelerometer signal and the gyro signal were fused using Kalman fusion sensor or
Kalman filter to get the angle value. Nowadays, a combo inertial measurement unit (IMU) consist of an
accelerometer and gyroscope is available at a very cheap price. PD controller calculates the appropriate
output for the motor using the angle and the rate angle information. All the calculations were done inside the
microcontroller. The scooter uses a PIC16F877 microcontroller. Figure 5 shows the controller box that is
consists of a PIC16F877 Microcontroller and a small battery to power the microcontroller, DC driver circuit,
and vertical gyro sensor. A 16x2 LCD is put on the box to show the PD parameters such as the proportional
gain and the differential gain. Outside the box, there are two potentiometers to set the PD gain and terminal
for serial communication. The microcontroller sends the PD output to the DC motor driver circuit. The DC
driver circuit uses the MOSFET IRF3205 H-Bridge circuit. The actuators of the scooter are two 250 watt DC
motors connected to two 13-inch wheelbarrow wheels. The two DC motors get the power from two 12 volt
dry cell batteries. Table 1 shows the list of the scooter's electronic components.
Figure 5. Controller box of the scooter
Table 1. Electronic components specification
No Function Components
1 Tilt sensor combo VG400 vertical gyro
2 Motor Driver MOSFET IRF3205 H Bridge
3 Actuator MY1060 250 watt DC Motor
4 Power Source Dry Cell Battery 12 V
2.3. Software design
The scooter will balance itself by moving the platform toward the leaning direction of the handlebar.
Information in which direction the system fall, how much the leaning angle now, and how fast the angle
changes between times are important. A proportional and derivative controller is selected to manage all those

 ISSN: 2088-8708
2458
information and calculate the torque needed to balance the system. The torque output for the motor is
calculated by multiplying the angle and the rate angle with some constants. The PD controller output was
converted into PWM value and sent to the DC motor driver. The DC motor will produce torque according to
the PWM signal sent to the DC motor driver. Both the left and right motors received the same amount of
value of PD output. However, the steering angle command differentiated the final control value sent to each
of the DC motors. The algorithm of the PD controller is shown:
𝑃 = 𝐾𝑝𝑒 (7)
𝐷 = 𝐾𝑑
𝑑𝑒
𝑑𝑡
(8)
𝑃𝐷 = 𝑃 + 𝐷 (9)
P is the value of the proportional controller, D is the derivative controller value, and PD is the
overall controller output. The 𝑒 is the error between the actual angle and the reference angle which is zero
degree. It tells how much the tilting angle of the scooter’s handlebar is. The
𝑑𝑒
𝑑𝑡
is the derivative of the error. It
tells how fast the scooter is handlebar falling backward or forward. The time cycle of the controller is 20 ms.
At a very small PWM value, the wheels will not move. It needs to add a constant from the output value.
However, the system needs a small dead band angle to improve the stabilization process by eliminating
chattering around the equilibrium.
2.3.1. Sensor fusion using Kalman filter
The output generation speed of the accelerometer signal is fast but full of noise. While the output of
the gyro signal is disturbed by gyro drift. Kalman Filter was used to fusing both of the signals to get better
angle reading. Kalman filter is a linear estimator. It estimates the states in the presence of process noise and
measurement noise. The process noise and the measurement noise are modeled as white noise [31]. Kalman
filter is popular in autonomous navigation research [32]. The mean squared error between the estimate and
the actual data is minimized. Consider the following equations:
State equation:
𝑥𝑘 = 𝐴𝑥𝑘−1 + 𝐵𝑢𝑘−1 + 𝑤𝑘−1 (10)
Output equation:
𝑧𝑘 = 𝐻𝑥𝑘 + 𝑣𝑘 (11)
Variable x is the state variable, and z is the measurement value with the measurement noise. In this
case, the state variables are output angle and gyro bias/drift. A and B are the matrices of the states The
process noise is represented by variable w, and the measurement noise is represented by variable v. Both of
the parameters are random variables and assumed to have a normal probability distribution. The covariances
Q and R also have a normal distribution and independent of each other. The covariance R can be determined
based on the real measurement variance. While Q can be determined intuitively. Based on the experiment, if
we put more on the Q accelerometer the output response is fast but noisy. If we put more on the Q gyro, the
output response is slower but less noisy. The best response for the tilt feedback is Q accelerometer=0.001 and
Q gyroscope=0.003. More detail on how to define matrices Q and R can be found in [33]. Kalman algorithm
predicts the state estimates. It has two continuous processes which are “priory” and “posteriori” state estimate
or “prediction” and “update” stages. The update is done by the sensor reading.
Initial error covariance and Initial estimated state must be described in the priory state estimate.
Other parameters that should be put in the Kalman filter loop are the noise covariance Ro and the process
noise covariance Qo. The predictor equations are as (12), (13):
𝑥
̂𝑘
−
= 𝐴𝑥
̂𝑘−1 + 𝐵𝑢𝑘−1 (12)
𝑃𝑘
−
= 𝐴𝑃𝑘−1𝐴𝑇
+ 𝑄 (13)
The (12) calculates state estimate of the state variable, and error covariance estimate is calculated in (13).
Error covariance is the error between the estimate of x and the true state. The following are the update
equations:

2459
𝐾𝑘 = 𝑃𝑘
−
𝐻𝑇
(𝐻𝑃𝑘
−
𝐻𝑇
+ 𝑅)−1
(14)
𝑥
̂𝑘 = 𝑥
̂𝑘
−
+ 𝐾𝑘(𝑧𝑘 − 𝐻𝑥
̂𝑘
−
) (15)
𝑃𝑘 = (𝐼 − 𝐾𝑘𝐻)𝑃𝑘
−
(16)
The error covariance P will be minimized by gain K, H and R are the measurement matrix and the
measurement covariance matrix respectively. The “a posteriori” state estimate is calculated in (15). Pk is the
covariance matrix of the “a posteriori”.
Both gyroscope and accelerometer have their drawbacks. The gyroscope signal tends to drift along
the time, and the accelerometer has significant measurement noise. To take only their advantages needs a
little bit of effort. Kalman fusion sensor is one of a good way for that. By combining some information, the
Kalman filter gets a good estimate of the actual tilt angle. Kalman filter was applied with angle and gyro bias
as the state variables. The steps of the Kalman Filter cycle is explained:
Step 1. Variable initialization
a) Determine the frequency of the gyro rate measurement to update the state.
b) Determine R (measurement covariance noise)
c) Determine R (process covariance noise)
Step 2. Routine updated every “dt”
a) Determine update of the covariance matrix
𝑃𝑘+1 = 𝑃𝑘+1 + 𝑃̇𝑑𝑡
𝑃̇ = 𝐴𝑃𝑘 + 𝑃𝑘𝐴′
+ 𝑄
b) Determine update of the estimated state
𝜃+= 𝑔𝑑𝑡
𝑔 = 𝑔𝑦𝑟𝑜_𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑚𝑒𝑛𝑡 − 𝑔𝑦𝑟𝑜_𝑏𝑖𝑎𝑠
Step 3. When the accelerometer reading is ready, the routine is updated.
a) Determine the Kalman gain 𝐾 = 𝑃𝑘+1𝐻𝑇
(𝐻𝑃𝑘+1𝐻𝑇
+ 𝑅)
b) Update the covariance matrix 𝑃𝑘+1 = (𝐼 − 𝐾𝐻)𝑃𝑘+1
c) Update the state estimate 𝑋+= 𝐾(𝑎𝑐𝑐𝑒𝑙𝑒𝑟𝑜𝑚𝑒𝑡𝑒𝑟_𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑚𝑒𝑛𝑡_𝑎𝑛𝑔𝑙𝑒 − θ))
2.3.2. Custom strategies for controlling the two wheels balancing scooter
A balancing scooter is a highly non-linear system. On the other hand, a PD controller is a linear
controller. Therefore, there are some non-linear problems on the system that a PD controller cannot handle.
We need to overcome those problems by adding some strategies to the controller. The biggest problem is
when the scooter cannot pull back the handlebar when the passenger sets its leaning angle too big by pushing
it too far. Secondly, a low-cost DC motor with a gearbox usually has a big backlash. A sudden change in the
direction of the motor rotation will cause a hard impact on the gearbox or the chain sprocket system. The
system’s response may oscillate and lead to instability. Figure 6(a) shows the main loop of the controller.
Besides the PD controller, there are exceeding angle routine, backlash routine, and steering routine.
Figure 6(b) shows the steering routine. Exceeding angle routine is used to limit the forward tilting angle. The
forward velocity is proportional to the leaning forward angle. The bigger the angle the faster the scooter will
run. The speed is limited by limiting the forward tilting angle to prevent over-speed. If the leaning angle is
too big, the system will push the handlebar backward. This also prevents the rider from falling forward if the
scooter cannot reach the required speed to tilt back the handlebar. The routine flowchart is shown in
Figure 6(c).
The scooter uses a sprocket and chain of a bicycle. It has a significant backlash that causes a
significant impact when the rotation direction of the motor is changed. If it happened near the equilibrium
position, it generates hardware vibration. It makes the mechanical part easy to be worn out. It also gives
significant disturbance to the balancing process. The routine of the anti-backlash is shown in Figure 6(d) it is
done by rotating the sprocket fast in just several milliseconds. Enough to fill the backlash gap and then
continue to follow the PD routine until the direction of the motor is changed again.

 ISSN: 2088-8708
2460
(a) (b)
(c) (d)
Figure 6. Flowchart of the scooter control algorithm (a) main controller loop, (b) steering algorithm, (c) push
handlebar back if exceeding max angle, and (d) anti backlash algorithm
3. RESULTS AND ANALYSIS
One method to tune the PID constant is Ziegler Nichols tuning method [34]. By using only
proportional controller, the scooter’s system response was plotted. The proportional gain was gradually
increased until the system response oscillates with a constant amplitude. The gain on that stage then called as
the ultimate gain, and the period of the oscillation called as the ultimate period. By using Ziegler Nicholes
table, the PID gain can be calculated. The ultimate oscillation due to the ultimate gain from the experiment is
shown in Figure 7. The achieved ultimate gain and the ultimate period are 6 and 1.088 second respectively.
Based on the Ziegler Nichols formula, the proportional gain is 3.6 and the derivative gain is 0.49.
Simulation result of the closed loop system from (6) is shown in Figure 8. The result shows that
using PD controller the scooter system can be balanced. Even when there is exist an impulse force on the
system, it can balance itself and move the scooter handlebar back in zero degree again. The scooter balancing
performance using the result from the Ziegler Nichols method is shown in Figure 9. The scooter can stay
balanced around the equilibrium or zero degree. But it oscillates with the amplitude in the average of
±2 degree.
The Ziegler Nichols, however, can give good starting value of the PD gain. But it needs manual fine
tuning to make the system more stable. By adding fine manual tuning, the best PD gains achieved are KP=3.6
and KD=5.39. KD is slightly higher because it is intended to reduce the overshoot. This combination of PD
gains satisfies the stability requirement of this system. The system response in Figure 10 shows an acceptable
result as the oscillation does not go beyond 1.5 degrees. The oscillation mostly happens at a negative angle.
It means that the handlebar a little bit leans in the forward direction. The negative angle of the system
indicates that the transporter leans to the front and makes an angle against the vertical direction. If the system

2461
tends to lean forward or backward, it can be corrected by adjusting the weight distribution of the parts or
components inside the system. If the angle is not so big, it can be corrected by changing the reference angle.
Changing the reference angle can fix the relative vertical position of the system. However, the platform might
be slightly inclined.
Figure 11 and Figure 12 show the inclination angle and the controller output signal to correct the
position without and with passenger respectively. The output response of the controller follows the algorithm
properly. On the application, the system limits the angle to 20 degrees. If the system angle is more than 20
degrees, the controller will not send any output to the motor. This is done to protect the rider so that the
system will not run too fast when the angle is too big.
Figure 7. The ultimate oscillation result from the
Ziegler Nichols PID tuning method
Figure 8. Simulation result of a scooter response to
an impulse input
Figure 9. The scooter balancing performance with
Ziegler Nichols PID gain tuning
Figure 10. The scooter balancing performance with
manual PD gain tuning.
Figure 11. Scooter without passenger balancing
performance
Figure 12. Scooter with passenger balancing
performance

 ISSN: 2088-8708
2462
Figure 13 shows the scooter testing on a tennis court. Figure 13(a) shows when the scooter at rest
and Figure 13(b) shows when the scooter has a rider on it. The scooter moved forward when the handlebar
was pushed forward. On the opposite, it runs backward when the handlebar is pulled backward. The bigger
the angle the faster the system will move.
The faster the motor is rotated, the lesser the torque is produced. This will affect the stability of the
system. Motor torque is reduced because the voltage from the battery is reduced by back EMF generated by
the motor itself. Moreover, if the scooter moves on a rough surface, it will experience larger disturbance. The
future recommendation is to study the variation of the passenger weight to the stability of the system.
Different inertia needs a different set of PID gain. Gain scheduling or PID adaptive can be a good candidate
for future investigation. Another approach is using a robust controller. H∞ robust controller with a simple
structure such as PID structure is also a good controller for a small embedded system [35]. With a robust
controller, the system is stable even in the presence of uncertainties including a variation of passengers.
(a) (b)
Figure 13. Hardware experiment on a tennis court (a) Scooter at rest position without passenger and
(b) Scooter cruises with passenger
4. CONCLUSION
The personal transporter or the balancing scooter is successfully developed. The scooter can be
ridden and controlled through the leaning handlebar and the steering potentiometer. A rider with maximum
65 kgs of weight successfully rides the scooter and does some maneuvers with it. Limitation of 65 kgs merely
due to the mechanical reason. The chassis that support the passenger is made from sheet metal. It starts to
deform when someone more than 65 kgs stands over it. The scooter has good maneuverability in limited
space. It can rotate 360 degrees in a spot. The system also has a good balance in the reverse direction. The
oscillation around the equilibrium takes less than plus-minus 1.5 degrees which is neglectable by the rider
since they will not realize it on the handlebar. When it has a rider on it, the oscillation is reduced due to the
damping effect from the rider’s hand.
REFERENCES
[1] H. J. T. Smith and J. A. Blackburn, “Experimental study of an inverted pendulum,” American Journal of Physics, vol. 60, no. 10,
pp. 909–911, Oct. 1992, doi: 10.1119/1.17012.
[2] O. Mayr, “The origins of feedback control,” Scientific American, vol. 223, no. 4, pp. 110–118, Oct. 1970, doi:
10.1038/scientificamerican1070-110.
[3] N. Yazdi, F. Ayazi, and K. Najafi, “Micromachined inertial sensors,” in Proceedings of the IEEE, 1998, vol. 86, no. 8,
pp. 1640–1659, doi: 10.1109/5.704269.
[4] M. R. M. Romlay, M. I. Azhar, S. F. Toha, and M. M. Rashid, “Two-wheel balancing robot; review on control methods and
experiments,” International Journal of Recent Technology and Engineering (IJRTE), vol. 7, no. 6S, pp. 106–112, 2019.
[5] F. Grasser, A. D’Arrigo, S. Colombi, and A. C. Rufer, “JOE: a mobile, inverted pendulum,” IEEE Transactions on Industrial
Electronics, vol. 49, no. 1, pp. 107–114, 2002, doi: 10.1109/41.982254.
[6] M. U. Draz, M. S. Ali, M. Majeed, U. Ejaz, and U. Izhar, “Segway electric vehicle,” in 2012 International Conference of Robotics
and Artificial Intelligence, Oct. 2012, pp. 34–39, doi: 10.1109/ICRAI.2012.6413423.
[7] T.-P. A. Perdicoulis and P. L. Dos Santos, “The secrets of segway revealed to students: revisiting the inverted pendulum,” in 2018
13th APCA International Conference on Control and Soft Computing (CONTROLO), Jun. 2018, pp. 43–48, doi:
10.1109/CONTROLO.2018.8514296.

2463
[8] K. Zakova and M. Hok, “Interactive three dimensional presentation of Segway laboratory model,” in 2016 International
Conference on Emerging eLearning Technologies and Applications (ICETA), Nov. 2016, pp. 377–380, doi:
10.1109/ICETA.2016.7802064.
[9] T. Kuwata, M. Tanaka, M. Wada, T. Umetani, and M. Ito, “Localization of segway RMP,” in SICE Annual Conference 2011,
2011, pp. 1675–1680.
[10] P. D. Ba, S.-G. Lee, S. Back, J. Kim, and M. K. Lee, “Balancing and translation control of a ball segway that a human can ride,”
in 2016 16th International Conference on Control, Automation and Systems (ICCAS), Oct. 2016, pp. 477–480, doi:
10.1109/ICCAS.2016.7832362.
[11] S. Back and S.-G. Lee, “Modeling and control system design of a ball segway,” in 2016 16th International Conference on
Control, Automation and Systems (ICCAS), Oct. 2016, pp. 481–483, doi: 10.1109/ICCAS.2016.7832363.
[12] Y. G. Rashid and A. M. A. Hussain, “Implementing optimization of PID controller for DC motor speed control,” Indonesian
Journal of Electrical Engineering and Computer Science (IJEECS), vol. 23, no. 2, pp. 657–664, Aug. 2021, doi:
10.11591/ijeecs.v23.i2.pp657-664.
[13] M. Abdulelah Alawan and O. J. Mohammed Al-Furaiji, “Numerous speeds-loads controller for DC-shunt motor based on PID
controller with on-line parameters tuning supported by genetic algorithm,” Indonesian Journal of Electrical Engineering and
Computer Science (IJEECS), vol. 21, no. 1, pp. 64–73, Jan. 2021, doi: 10.11591/ijeecs.v21.i1.pp64-73.
[14] S. N. Al-Bargothi, G. M. Qaryouti, and Q. M. Jaber, “Speed control of DC motor using conventional and adaptive PID
controllers,” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 16, no. 3, pp. 1221–1228, Dec.
2019, doi: 10.11591/ijeecs.v16.i3.pp1221-1228.
[15] O. A. Trujillo, N. Toro-García, and F. E. Hoyos, “PID controller using rapid control prototyping techniques,” International
Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 3, pp. 1645–1655, Jun. 2019, doi:
10.11591/ijece.v9i3.pp1645-1655.
[16] M. H. Salehpour, H. D. Taghirad, and H. Moradi, “Two PID-based controllers for a tethered segway on dome shaped structures,”
in 2018 6th RSI International Conference on Robotics and Mechatronics (IcRoM), Oct. 2018, pp. 577–582, doi:
10.1109/ICRoM.2018.8657643.
[17] Y. Ye, C.-B. Yin, Y. Gong, and J. Zhou, “Position control of nonlinear hydraulic system using an improved PSO based PID
controller,” Mechanical Systems and Signal Processing, vol. 83, pp. 241–259, Jan. 2017, doi: 10.1016/j.ymssp.2016.06.010.
[18] A. Kaba, “Improved PID rate control of a quadrotor with a convexity-based surrogated model,” Aircraft Engineering and
Aerospace Technology, vol. 93, no. 8, pp. 1287–1301, Sep. 2021, doi: 10.1108/AEAT-03-2021-0093.
[19] M. Rahman, Z. C. Ong, W. T. Chong, and S. Julai, “Smart semi-active PID-ACO control strategy for tower vibration reduction in
wind turbines with MR damper,” Earthquake Engineering and Engineering Vibration, vol. 18, no. 4, pp. 887–902, Oct. 2019, doi:
10.1007/s11803-019-0541-6.
[20] B. Sumantri, E. H. Binugroho, I. M. Putra, and R. Rokhana, “Fuzzy-PID controller for an energy efficient personal vehicle: Two-
wheel electric skateboard,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 6, pp. 5312–5320,
Dec. 2019, doi: 10.11591/ijece.v9i6.pp5312-5320.
[21] T. Qi, C. Li, W. Kai, and L. Yan, “Research on motion control of mobile robot with fuzzy PID arithmetic,” in 2009 9th
International Conference on Electronic Measurement and Instruments, Aug. 2009, pp. 3–366, doi:
10.1109/ICEMI.2009.5274284.
[22] Á. Odry, R. Fullér, I. J. Rudas, and P. Odry, “Fuzzy control of self‐balancing robots: A control laboratory project,” Computer
Applications in Engineering Education, vol. 28, no. 3, pp. 512–535, May 2020, doi: 10.1002/cae.22219.
[23] C.-F. Hsu and W.-F. Kao, “Double-loop fuzzy motion control with CoG supervisor for two-wheeled self-balancing assistant
robots,” International Journal of Dynamics and Control, vol. 8, no. 3, pp. 851–866, Sep. 2020, doi: 10.1007/s40435-020-00617-y.
[24] P. Sutyasadi and M. Parnichkun, “Push recovery control of quadruped robot using particle swarm optimization based structure
specified mixed sensitivity H2 / H∞ control,” Industrial Robot: the international journal of robotics research and application,
vol. 47, no. 3, pp. 423–434, Mar. 2020, doi: 10.1108/IR-06-2019-0135.
[25] S. Chantarachit, “Development and control segway by LQR adjustable gain,” in 2019 International Conference on Information
and Communications Technology (ICOIACT), Jul. 2019, pp. 649–653, doi: 10.1109/ICOIACT46704.2019.8938489.
[26] R. Babazadeh, A. G. Khiabani, and H. Azmi, “Optimal control of Segway personal transporter,” in 2016 4th International
Conference on Control, Instrumentation, and Automation (ICCIA), Jan. 2016, pp. 18–22, doi: 10.1109/ICCIAutom.2016.7483129.
[27] W. Netto, R. Lakhani, and S. M. Sundaram, “Design and performance comparison of different adaptive control schemes for pitch
angle control in a twin-rotor-MIMO-system,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9,
no. 5, p. 4114, Oct. 2019, doi: 10.11591/ijece.v9i5.pp4114-4129.
[28] A. Odry and R. Fuller, “Comparison of optimized pid and fuzzy control strategies on a mobile pendulum robot,” in 2018 IEEE
12th International Symposium on Applied Computational Intelligence and Informatics (SACI), May 2018, pp. 207–212, doi:
10.1109/SACI.2018.8440947.
[29] A. Chhotray and D. R. Parhi, “Navigational control analysis of two-wheeled self-balancing robot in an unknown terrain using
back-propagation neural network integrated modified DAYANI approach,” Robotica, vol. 37, no. 8, pp. 1346–1362, Aug. 2019,
doi: 10.1017/S0263574718001558.
[30] R. C. Ooi, “Balancing a two-wheeled autonomous robot,” University of Western Australia School of Mechanical Engineering,
2003.
[31] O. Bayasli and H. Salhi, “The cubic root unscented kalman filter to estimate the position and orientation of mobile robot
trajectory,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 5, pp. 5243–5250, Oct. 2020, doi:
10.11591/ijece.v10i5.pp5243-5250.
[32] L. Lasmadi, A. I. Cahyadi, S. Herdjunanto, and R. Hidayat, “Inertial navigation for quadrotor using Kalman filter with drift
compensation,” International Journal of Electrical and Computer Engineering (IJECE), vol. 7, no. 5, pp. 2596–2604, Oct. 2017,
doi: 10.11591/ijece.v7i5.pp2596-2604.
[33] Y. Zahraoui and M. Akherraz, “Kalman filtering applied to induction motor state estimation,” in Dynamic Data Assimilation-
Beating the Uncertainties, IntechOpen, 2020.
[34] M. Nafea, A. R. M. Ali, J. Baliah, and M. S. Mohamed Ali, “Metamodel-based optimization of a PID controller parameters for a
coupled-tank system,” Telecommunication Computing Electronics and Control (TELKOMNIKA), vol. 16, no. 4, pp. 1590–1596,
Aug. 2018, doi: 10.12928/telkomnika.v16i4.9069.
[35] P. Sutyasadi and M. B. Wicaksono, “Joint control of a robotic arm using particle swarm optimization based H2/H∞ robust control
on arduino,” Telecommunication Computing Electronics and Control (TELKOMNIKA), vol. 18, no. 2, pp. 1021–1029, Apr. 2020,
doi: 10.12928/telkomnika.v18i2.14749.

 ISSN: 2088-8708
2464
BIOGRAPHIES OF AUTHORS
Petrus Sutyasadi earned his master degree in mechatronics in 2008 and
doctoral degree in mechatronics in 2016 from Asian Institute of Technology, Thailand.
Currently he is a lecturer at Mechatronics Department in Sanata Dharma University,
Yogyakarta, Indonesia. His research interest includes robust control, robotic, and frugal
innovation in mechatronic engineering. He can be contacted at email: peter@usd.ac.id.
Manukid Parnichkun earned his master degree in Precision Machinery
Engineering in 1993 and doctoral degree in mechatronics in 1996 from The University of
Tokyo, Japan. Currently he is a full professor at Industrial System Engineering Program in
The Asian Institute of Technology, Bangkok, Thailand. His research interest includes new
robot mechanism, novel control algorithm, and innovative measurement concept. He can be
contacted at email: manukid@ait.asia.

Developing low-cost two wheels balancing scooter using proportional derivative controller

More Related Content

Similar to Developing low-cost two wheels balancing scooter using proportional derivative controller (20)

More from IJECEIAES (20)

Recently uploaded (20)

Developing low-cost two wheels balancing scooter using proportional derivative controller