A Fuzzy Integrated Self-tuning PID Technique for

Mini Robot

B.G. Sampath1, G.V.A.G. Asanka Perera2 ,W.D.I.G Dassanayake3

Department of Mechanical Engineering1,3, Department of Electrical Engineering2
University of Moratuwa
Katubedda, Moratuwa, Sri Lanka

Abstract— This paper presents the modeling, numerical accurate trajectories in the presence of the noise and time
simulation, control and the testing of a fuzzy integrated self- variant system.
tuning PID controller for a line following mini robot. The system
is modeled for different lighting conditions considering the Combining PID controllers with fuzzy and genetic
desired and ambient light intensities. The system modeling is algorithms (GA) [3] can provide effective solutions for
performed in a simulation environment and the generated sophisticated applications. Fuzzy and GA based controller for
control parameters are applied to the tuned fuzzy controller. The an antilock braking system (ABS) was studied by Sharkawy in
proposed controller architecture is extended to achieve smooth [4]. The proposed architecture was applied to ABS with a self-
and fast movements of the robot under varying ambient and tuning PID controller. The significant aim of this study was to
surface conditions. This extension of the fuzzy based PID improve the performance of ABS application to withstand the
controller can be applied to similar applications which deal with system nonlinearities and other adverse effects caused by
high degrees of environmental lighting changes. The tuned weight, friction (static and viscous) of the road and road
system is capable of self-tune the PID parameters without human inclinations.
interaction. The proposed design shows a significantly improved
performance over its conventional counterpart. In the context of self-tuning controllers, Woo et al. [5]
introduced a PID type fuzzy controller with self-tuning scaling
Keywords— Self-tuning PID, fuzzy logic controller, line factors to improve the performance of the transient and steady
following robot state system responses. Further improvements of PID and
fuzzy integrated controllers can be found in [6-10].
I. INTRODUCTION In this paper, the capabilities of a self-tuning fuzzy
Modern industrial robots and automated guided vehicles integrated PID controller are extended towards a line following
(AGVs) play a vital role in manufacturing and material mini robot application. The mini robot’s trajectory tracking and
handling processes. Their assistance is integrated at different decision making skills for varying ambient conditions are
stages ranging from raw materials intake to assembly line greatly improved with the proposed fuzzy based algorithm.
handling at different manufacturing scales, and finally at the Although, the fuzzy based PID controllers are already well
end of the production line, to perform storage management at established area, this paper contributes with an extended
where houses. controller design for an ambient lighting conditioned and
varying surface (reflection and color characteristics)
Most of the control systems of line following AGVs and application. The algorithms designed for these conditions can
Security robots are tuned using proportional-integrative- be modified and applied to any similar application which deals
derivative (PID) control techniques and they are manually with environmental condition based decision making.
tuned and tested, which is considered as the traditional method
of tuning PID parameters. Since this is a trial and error method, The proposed methodology consists of modeling,
it is not the best practice to tune a PID controller. Self- numerical simulation, controlling and the testing of a fuzzy
organizing controllers such as fuzzy integrated controllers integrated self-tuning PID controller for the line following mini
extend the traditional PID controllers’ capabilities with an robot. Further, the system is modeled for two basic control
added learning ability. inputs: deviation error from the desired line and ambient light
intensity. The system model is obtained by deriving the
Various studies have been proposed for fuzzy based transfer equation of motion and the system variables generated
controllers. Fuzzy PID controlled shape memory alloy (SMA) with Matlab Simulink tool. Then, the PID parameters are
actuator was introduced by Zhenyum et al. [1] for a two DOF attained by using tuned fuzzy controller. In order to perform
joint. In this design, they designed a sensorless SMA servo these formulations, fuzzy membership functions are defined for
control using polynomial functions, and integrated fuzzy PID the two inputs, deviation error from desired line and ambient
controllers to minimize the overstress while increasing the light intensity. The system is capable of self-tune PID
response time. A conceptually similar approach was proposed parameters without human interaction, and ensures smooth and
for a two-link non-linear revolute-joint robot arm by Kazemian fast movement of the robot in different ambient conditions,
[2]. The proposed self-tuning architecture was able to follow

978-1-4799-4598-6/14/$31.00 ©2014 IEEE

This paper is categorized as follows. In the section II, the reflection and diffuse reflection. In this study, two surfaces are
methodology used to develop the fuzzy integrated controller arranged to satisfy the specular reflection and diffuse
and the theoretical backgrounds are explained. Further, the reflection, and obtained the results for the ambient lighting
proposed control algorithm and its designing steps are conditions. The ambient light intensity which affects the PID
explained. Section III describes the experimental setup, and tuning is varied in the experimental set up to determine results.
section two discusses the results and compares them with
conventional controllers based system responses. Finally, the A. Control system design
paper is concluded in section IV, The fuzzy integrated self-tuning PID controller is an auto-
tuning controller designed by incremental fuzzy logic
II. METHODOLOGY controller in place of the proportional term in traditional PID
The line follower has to identify two zones in the moving controller. This arrangement fine tunes the parameters of PID
arena which are white line and black background as shown in controller on line by fuzzy rules. The time dependent inputs are
Fig. 1. The feedback process is performed by an infrared error e and rate change of error e were used as control
analog sensor panel mounted in front of the robot. The output inputs for the fuzzy controller.
signal depends on orientation of robot with respective to the
white line and the noise. The noise is caused by the ambient de(t)
light and nature of the black and white surfaces. At the desired
orientation of the robot, it generates a 2.5V signal at the sensor
³
U(t) = K P .e(t) + K I . e(t).dt + K D .
dt
(1)

panel output. An additional sensor module is installed to filter

out the disturbance conditions. However, the sensor panel Where, K P , K I and K D are proportional, integral and
output varies depending on the environmental conditions. derivative gains respectively [5,11]. These gains also have
According to the sensed signal, the PID parameters are following meanings:
analyzed using fuzzy controller.
K PT K T
KD = , KD = P D (2)
Ti T

T is the sample time and Ti is the integral time parameter.

When integrating the proposed controller, there is no need
to redesign existing hardware, and only the controlling
algorithms have to be modified. Fuzzy integrated system
analyzes fuzzy relations between PID parameters and
minimizes the overshoot, settling time and steady state error.
There are many studies on determining the parameters and
updating them according to the dynamic time varying
situations [5]. Combinations of each PID parameter values are
considered in this study to improve the performance. Table I
shows the effects of the increment of each parameter on the
system behavior.

Fig. 1. Robot orientation TABLE I. THE EFFECTS OF INCREASING PID PARAMETERS ON THE
SYSTEM BEHAVIOUR

The line following mini robot is designed to detect a 30mm Steady state
Rise time Overshoot Settling time
white line pasted on a black surface using infrared sensors. error
When the mini-robot is moving following the line, the self- KP Decreased Increased Small change Decreased
tuning process will get affected by the ambient lighting KI Decreased Increased Increased Eliminate
conditions; because lighting conditions in the environment add KD Small change Decreased Decreased No change
as a noise to the control system through infrared sensors which
detect the line. The ambient lighting conditions can vary from
daylight to night (natural lighting conditions), and in some Proportional gain coefficient reduces the system deviation
special cases artificial lighting will be present in the from the desired orientation. If K P is increased, it results in
environment. However, the ambient light condition depends on
an increased response speed and steady state error also closes
the amount of light reflected from the surface. Thus, the
infrared sensors which measure the reflected light from the to zero. If K P is a much higher value, then the system
surface of the line depend on the ambient light intensity and becomes unstable. Conversely, a small K P reduces overshoot
surficial reflectivity. and increase the system stability margin while deteriorating
Based on the surficial reflexivity, the reflection of light the precision.
from the surfaces can be divided into two types as specular
Integral gain of the system eliminates steady state error, but TABLE IV. THE FUZZY RULE SET FOR KD
it makes system response slow down, and generates KD NL NS Z PS PL
oscillations with an increased overshoot. If K I is increased, NL Z Z PS PS PL
then system biases to reduce the steady state error. However, NS Z Z Z Z PS
when K I is too large, overshoot becomes significant, and Z Z Z Z PS PL
system may produce a oscillating behavior. Stability of the PS PS PS PS PL Z
system can be reached with a decreasing K I . PL Z Z Z PS PL
Differential gain is reflecting the changing nature of
deviation of the system. When the deviation signal changes
largely, it can introduce an effective signal at early stages, and Fuzzy rule matrixes for each three parameters and
it is propitious to reduce the overshoot, overcome the system membership function for inputs and output parameters are
oscillation, and makes the system stable. If KD is too large shown from Table II-IV. Input and output membership
then the response process will be in advance braking, thereby functions are given in Fig. 2 and 4, respectively. Parameter
the adjusting time will be delayed. If K D is very small, then estimation is very complex and challenging task, and with the
conventional methods, it is difficult to realize the optimum
the decelerating of system adjusting time will be delayed, parameter quantities. Self-tuning algorithm always analyzes
resulting an increased overshoot, slow response time and an and modifies the control parameters automatically even at the
unstable system. working stage. Once system performance is changed beyond
The PID parameters are changing simultaneously according the control limits of the system, the controller can identify it
to the input signal and output of the system. The practical and return to the optimum working condition without an
experiments observed that the values of these parameters are external involvement. Therefore, this proposed controller can
changing around 0.5 to 2.5 in different ambient conditions. In improve real time system performances operating under
this study, the tuned values for given conditions are saved to different environment condition.
improve fast tuning when the system operates in different
environmental conditions. The fuzzy rules are defined using
Mamdani’s fuzzy interference method with two inputs are e
and e , and outputs are ΔK P , ΔK I and ΔK D .

K P + 1 = K P ± ΔK P (3)

K I + 1 = K I ± ΔK I (4)

K D + 1 = K D ± ΔK D (5)

TABLE II. THE FUZZY RULE SET FOR KP

NL NS Z PS PL
NL PVL PVL PVL PL PM
NS PVL PVL PVL PL PM
Z PL PL PM PS PS
PS PM PS PS PS PS
PL PS PS Z Z Z
Fig. 2. Block diagram for fuzzy adaptive PID controller

TABLE III. THE FUZZY RULE SET FOR KI

KI NL NS Z PS PL
NL PVL PL PM PM PM
NS PVL PL PL PM P
Z PM PS Z Z Z
PS PM PM PS Z Z
PL PS Z Z Z Z
Fig. 3. Input membership functions
self-tuned according to the control algorithm. All gain
parameters reach to the steady state with a lesser overshoot.

Fig. 4. Output membership functions

III. EXPERIMENTAL SETUP

The robot consists of two motors with two wheels and
single caster wheel (Fig. 5). The testing environment with
different lighting and surface conditions are showed in Fig. 6.
When considering the hardware parameters of the robot, it is
very important to configure the sensors and filtering
techniques. Five infrared LEDs and photo transistor couples
are used as sensing elements for each side and sensor panel has
been design the way that it gives single analog signal according
to orientation of robot with respective to the white line. Two
12V DC geared motor (Gear ratio 1:20 and motor speed 12000
rpm) and single caster wheel are fixed to the body frame in
symmetric triangular manner to make ease of maneuverability.
Motor speeds are controlled by 120 KHz PWM signal
according to control system.

Fig. 7. Variation of self-tuning PID gain values under fixed lighting

conditions. (a) Kp variation. (b) KI variation. (d) KD variation.

Fig. 8 compares the output responses between fuzzy

integrated PID controlled system with a conventional PID
controlled system. The results show that when the fuzzy based
self-tuning is combined, the controller’s performance improves
significantly with reduced overshoot and low settling time.
Fig. 5. Mini-robot platform

Fig. 6. Testing environment with different lighting and surface conditions

IV. RESULTS
Fig. 8. Output tuning diagram of fuzzy integrated PID contoller vs
The initial PID parameters are obtained using a simulation conventional PID contoller.
model and they are further tuned with trial and error method.
The PID gain parameters are self-tuned under a fixed lighting The sensor panel (Fig. 1) output for two different lighting
condition. Fig. 7 (a)-(c) show their variation over time, when conditions are drawn in Fig. 9. For ideal environmental lighting
conditions, the graphs should follow straight lines. However,
due to noises caused by the dynamic environment and system
non-linearities, the sensed signals deteriorate from their ideal
shape. These variations effect on the control parameters of the
system. The real time experimental output graphs are given
from Fig. 10-13 for 400ms interval.
Fig. 10 gives the self-tuning of PID gain values. When the
ambient light condition changed from the initial condition to a
different lighting level the PID gain values also change. This
scenario is represented in Fig. 11. A similar environmental
model is shown in Fig. 6. In the PID only method, the
overshoot remains for a considerable period of time, while the
proposed controller based system output settles very quickly
with a small overshoot.
The system responses of the PID only controller and the
fuzzy integrated PID controller are compared in Fig. 12 for
constant ambient conditions. The system response has
improved in terms of overshoot and settling time. At the Fig. 9. Sensor panel output for different ambient conditions
ambient condition changing zone, the overshoot is compared
for PID only and Fuzzy PID conditions. The proposed system
shows comparatively better performance when operating under
changing lighting and surface conditions (Fig. 13).

Fig. 10. Self-tuned KP, KI and KD variations (a) KP variation (b) KI variation (c) KD variation

Fig. 11. Change of KP, KI and KD at the ambient light condition changing zone. (a) KP variation (b) KI variation (c) KD variation
Fig. 112. (a) PID Output tuning diagram (b) Fuzzy PID Output tuning diagram

Fig. 12. (a) PID Overshoot comparison at ambient condition changing zone (b) Fuzzy PID Overshoot comparison at ambient
condition changing zone

application. This approach particularly applicable to a

V. CONCLUSION autonomous vehicle operating under varying lighting and
Considering the variation of tuned PID parameters for same surface conditions. The algorithms designed for these
system under different ambient conditions, this study has conditions can be modified and applied to any similar
shown the modeling, simulation and testing of self-tuning PID application which deals with environmental condition based
controller for mini line following robot. The increased system decision making.
performance of the proposed method is validated with the
system response comparisons with conventional PID only VI. REFERENCES
approach. All PID parameters are analyzed by fuzzy adaptive [1] Zhenyun Shi, Tianmiao Wang, Da Liu, Chen Ma, Xiangnan Yuan, “A
controller to minimize overshoot, settling time and steady state fuzzy PID-controlled SMA actuator for a two-DOF joint”, Chinese
error. Journal of Aeronautics, Volume 27, Issue 2, April 2014, pp 453-460
[2] Kazemian, Hassan B. "Comparative study of a learning fuzzy PID
The design is dependent on the experience and knowledge controller and a self-tuning controller." ISA transactions 40.3 (2001):
of the designer. Although PID parameters are automatically 245-253.
self-tuned, defining fuzzy rules is a difficult task. The obtained [3] Chiang, Chih-Knan, Hung-Yuan Chung, and Jin-Jye Lin. "A self-
results show that self-tuned PID fuzzy integrated controller is learning fuzzy logic controller using genetic algorithms with
able to reach the intended performances reinforcements." Fuzzy Systems, IEEE Transactions on 5.3 (1997): 460-
467.
In this study, the capabilities of a self-tuning fuzzy [4] Badie Sharkawy, Abdel. "Genetic fuzzy self-tuning PID controllers for
integrated PID controller are extended towards a line following antilock braking systems." Engineering Applications of Artificial
mini robot application. The mini robot’s trajectory tracking and Intelligence 23.7 (2010): 1041-1052.
decision making skills for varying ambient conditions are [5] Woo, Zhi-Wei, Hung-Yuan Chung, and Jin-Jye Lin. "A PID type fuzzy
greatly improved with the proposed fuzzy based algorithm. controller with self-tuning scaling factors." Fuzzy Sets and Systems
115.2 (2000): 321-326.
Although, the fuzzy based PID controllers are already well
established area, this paper contributes with an extended [6] Mudi, Rajani K., and Nikhil R. Pal. "A robust self-tuning scheme for PI-
and PD-type fuzzy controllers." Fuzzy Systems, IEEE Transactions on
controller design for an ambient lighting conditioned and 7.1 (1999): 2-16.
varying surface (reflection and color characteristics)
[7] Jiang, Wei, and Xuchu Jiang. "Design of an Intelligent Temperature
Control System Based on the Fuzzy Self-Tuning PID." Procedia
Engineering 43 (2012): 307-311.
[8] Soyguder, Servet, Mehmet Karakose, and Hasan Alli. "Design and
simulation of self-tuning PID-type fuzzy adaptive control for an expert
HVAC system." Expert Systems with Applications 36.3 (2009): 4566-
4573.
[9] Lee, Chuen-Chien. "Fuzzy logic in control systems: fuzzy logic
controller. II." Systems, Man and Cybernetics, IEEE Transactions on
20.2 (1990): 419-435.
[10] Sagirli, Ahmet, et al. "Self-tuning fuzzy logic control of crane structures
against earthquake induced vibration." Nonlinear Dynamics 64.4 (2011):
375-384.
[11] Visioli, A. "Optimal tuning of PID controllers for integral and unstable
processes." IEE Proceedings-Control Theory and Applications 148.2
(2001): 180-184.