Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/351051759

A Review on Intelligent PID Controllers in Autonomous Vehicle

Chapter · April 2021


DOI: 10.1007/978-981-15-7675-1_39

CITATIONS READS

4 609

5 authors, including:

Vagisha Vartika Subhranil Das


Birla Institute of Technology, Mesra MIT World Peace University
1 PUBLICATION 4 CITATIONS 23 PUBLICATIONS 38 CITATIONS

SEE PROFILE SEE PROFILE

Sudhansu Mishra Sitanshu Sekhar Sahu


Birla Institute of Technology, Mesra Birla Institute of Technology, Mesra
108 PUBLICATIONS 1,023 CITATIONS 82 PUBLICATIONS 1,534 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Subhranil Das on 10 June 2021.

The user has requested enhancement of the downloaded file.


A Review on Intelligent PID Controllers in
Autonomous Vehicle
Vagisha Vartika 1, Swati Singh 1, Subhranil Das 1, Sudhansu Kumar Mishra 1, and Sitanshu Sekhar
Sahu2
1
Control Systems, Electrical and electronics, BIT Mesra, Ranchi
{vagishavartika,swatisingh07}@gmail.com,
sudhansumishra@bitmesra.ac.in
2
Control Systems, Electrical and electronics, BIT Mesra, Ranchi
subhranil.sdas2007@gmail.com

Abstract. In recent times, autonomous vehicle is gaining popularity in


several of applications ranging from conventional transport system to nuclear
reactor. The importance of autonomous vehicle increases manifold for those
applications where it is not possible or very much risky for human to reach
out. Many Artificial Intelligence techniques have been successfully applied
in this field. In this paper an adequate number of machine learning based
techniques for designing intelligent PID controllers in autonomous vehicles
have been reviewed and some of the recent state-of-the art techniques have
been listed and suitably compared among them. Particularly, the parameters
tuning of Kp, Ki, Kd as well as the rise time characteristics by two methods
namely Ziegler- Nichols and genetic Algorithm (GA) have been discussed.
Also, we have compared the properties of rise time and percentage
overshoot by conventional PID Controllers and fuzzy based PID controllers.
Keywords: Genetic algorithm, fuzzy logic, Ziegler Nichols, autonomous ground vehicle.

1 Introduction

An autonomous vehicle is an autonomous robot which operates without any human


onboard, by applications of Artificial Intelligence techniques [1]. Autonomous
vehicles can be used for those applications where humans cannot reach, or maybe
it is very inconvenient and dangerous for humans to reach out [2]. With the
support of governments and cooperation of information technology sector, recent
advances have been made in the area of robotics. Recently, C.Li et.al. have
proposed an algorithm which is based on the challenges that deals with reference
path for Autonomous Ground Vehicles(AGV). The Clothoid curve method has
been applied as the path primitive to apt the remaining part of reference path [3].
The application of Autonomous vehicles (AV) has been greatly expanded in recent
decade in the areas, such as, industrial, mining, natural environment, area mapping,
scene manipulation and understanding etc. Combination of mapping and
localization methods as well as obstacle detection and avoidance could be
translated into some useful platforms which can increase safety of products in
smart industries [4]. Sensors have been embedded in the vehicle so as to develop
understanding of the environment and used by some control algorithms for the
determination of next action. The autonomous robots can carry out tasks, such as
assisting friendly troops and bomb defusing in the fields of military operations [1].
2

Variety of research work has been done for identification of the shortest path like
Lazy Theta*. Development of integrated network environment (INE) for robotics
system which is based on internet because all the robotics systems is actually
isolated by different environments from one another and there is no effective way
for communication [1]. However, AV is not mature enough to operate in the
complex environment, and significant research is yet to come. The path planner is
the vital part of AV, but still has many problems which are not being solved.

2 Tuning of PID Controllers in Autonomous Ground Vehicles


(AGV)

The concept of developing PID Controllers had been started in 1911 by Elmer
Sperry [2]. After 2 decades, Taylor Instrumental Company (TIC) introduced the
first pneumatic controller with a fully tunable proportional controller. Thereafter,
control engineers devised first PID pneumatic controller with derivative action in
order to eliminate steady state error. In the year of 1942, Ziegler and Nichols
proposed tuning rules so as to identify the parameters of PID Controllers.

The typical PID control in standard form is written as

𝑑𝑒(𝑡) 1 Ω
u(t)=Kp[e(t)+T 𝑑𝑡
+ Ti ∫0 𝑒(𝑡)𝑑𝑡] (1)

where e(t) =y’(t) - y(t)is the system error (difference between the reference input
and the system output), u(t) the control variable, Kp the proportional gain, Td the
derivative time constant and Ti the integral time constant. Eqn. 1 can also be written
as

𝑑𝑒(𝑡) Ω
u(t)=Kp e(t)+ Kd 𝑑𝑡
+ Ki ∫0 𝑒(𝑡)𝑑𝑡 (2)

where Kd=KpTd and Ki=Kp/Ti. The tuning problem consists of determining the
values of these three parameters with the aim of satisfying different control
specifications, such as set-point following, load disturbance attenuation, robustness
to model uncertainties and rejection of measurement noise [2].Industrial
applications of PID controllers began in mid-1950s.For the robust operation of PID
Controllers, Internal Model Control (IMC) tuning methods have been presented for
large number of classes of applications in industry. Controller tuning can be
effortlessly executed by adjustment of time constant of filter in First Order plus
Delay Time (FOPDT) Model [2].

PID controllers are ideally suitable for first and second order equations. The
implementation of PID controllers in industries involve higher order set of
equations where tuning becomes necessary condition. After the introduction of
Artificial Intelligence (AI) techniques with the combination of PID controllers have
3

made enormous step on controlling the tuning methods as compared to Ziegler-


Nicholas method (ZN). One of the applications of PID Controller, the selection of
Kp, Ki and Kd becomes crucial condition which is based on tuning methods, such as
Empirical Ziegler-Nichol formula and Modified Zeigler- Nichol formula [5].

2.1 Ziegler-Nichols first method of tuning

The transfer function C(s)/U(s) can be approximated by a first order system with a
lag as follows:

𝐶(𝑠) 𝐾 𝑒 _𝐿𝑠
𝑈(𝑠)
= 𝑇𝑠+1 (3)

ZN method suggested a set of values of Kp, Ti and Td according to table-1.

Table 1. ZN first method of tuning

Type of Kp Ti Td
controller
P 𝑇 ∞ 0
𝐿
PI 0.9
𝑇 𝐿 0
𝐿
0.3
PID 𝑇 2L 0.5L
1.2 𝐿

2.2 Ziegler Nichols second method of tuning

The values of Kp, Ti and Td are determined from critical stability and critical gain.
Kp, Ti and Td can be determined from the values mentioned in Table-2.

Table 2. ZN first method of tuning

Type of Kp Ti Td
controller
P 0.5Kcr ∞ 0
PI 0.45Kcr 1 0
Pcr
1.2

PID 0.6Kcr 0.5 Pcr 0.125 Pcr


4

The disadvantage of ZN method is the result does not provide satisfactory


performance every time. The concept of Fuzzy Logic (FL) has been introduced to
search out for the desired performance so as to rectify the drawbacks of ZN method.
The parameters of PID Controller can be modified according to the plant need with
the help of new techniques, such as Fuzzy Logic Controller (FLC), Genetic
algorithm (GA) etc. which has been discussed in detail in subsequent sections.

3 Fuzzy Logic

Fuzzy logic (FL) is an attempt to mimic the human reasoning process that can be
used to identify and control complicated systems. Fuzzy logic Control (FLC) has
the advantage of incorporating human knowledge for making PID as an adaptive
controller [6]. Fuzzy PID controller can be self-tuned which can automatically
adjust the parameters of Kp, Ki, Kd in order to increase the performance by means
of with or without fuzzy tuning of nonlinear systems [6].The output of fuzzy set
from the error and derivation error is used to tune the PID parameters with
appropriate value [3]. Recently, an improved fuzzy based PID Controller have been
designed by adding more control laws to fuzzy PI as well as fuzzy PD which has
been shown in below figure1[1].

Fig.1: Block diagram of FLC tuned PID

The abbreviations used in Fuzzy Inference system are as follows:


NVB --- negative very big
NB --- negative big
NM --- negative medium
NS --- negative small
Z --- zero
PS --- positive small
PM --- positive medium
PB --- positive big
5

PVB --- positive very big

FLC technique guarantees good performance in set point and load disturbance step
response, making this technique promising for PID tuning implementations

4 Implementation of Genetic Algorithm based tuning of PID


Controller

As compared to conventional methods of tunable PID Controllers, Genetic


Algorithms (GA) have resulted improved performance. Darwin’s principle forms
the basis of GA and requires no pre-existing knowledge of initial conditions of given
system. The searching process considered as a biological process where stronger
individuals are likely to be the winner in a competitive environment. To initialize
the GA, a set of solutions are represented by chromosomes known as population.
Each solution can be represented by a binary string of ones and zeros, real number
or in other forms that depends upon the characteristic of application data. In such
algorithms, the fittest individual survive among a group of individuals. This fittest
individual forms a new generation of individuals with more improved fitness values.
The measure of fitness of an individual depends upon the performance in the
problem domain. In last few decades, GA has been recognized as an effective and
efficient technique to solve optimization problems. GA begin with an initial
population consisting of a number of chromosomes where each chromosome
represents a solution to the problem evaluated by a fitness function.
Basically, GA works on three operating conditions namely, selection, mutation and
crossover. Selection provides better chances of reproduction to the fittest individual.
During crossover, some reproduced individuals exchange their genetic
characteristics. Mutations occur in a diminutive percentage which have resulted in
a random change of a genetic material. The evolution process guides the population
through more promising regions in the search space. After performing of these three
operations, new set of individual are being created which has been proved to be
better than parents. This algorithm finally stops when the optimum solution is
reached [7]. The implementation of the tuning procedure through GA starts with the
definition of the chromosome representation where Kp,Ki, and Kd are represented
by three individual chromosomes which is given in the figure2.

Fig.2: Chromosome definition

Here, the three gains are considered to be binary string. Each gain have been
represented by a ten binary numbers making each chromosome having thirty genes
which is given in the figure 3.
6

Fig.3: GA based PID Controller

4.1 Steps of Classic Genetic Algorithm (CGA) based PID Tuning

1) Binary encoding have been adopted where each binary cluster has one value
between 0-1024(210). The approximate length of each parameter is being
determined and encoded.
2) Initial random population, N has been created. The gene expression k=[kp,ki,kd]
coded in binary number to increase the efficiency and improve the calculation speed
and robustness.
3) Each population member has been decoded and fitness values for each individual
n are saved under current population N. The fitness function is defined as

J=∫ (w1|e(t)| + w2u2(t)) dt + w3tu (4)
0

where, e(t) is the error between the reference signal and the system output is, u(t) is
the output of controller, tu is the rising time, wi (i=1,2,3) are the corresponding
weight values. A penalty function is added to objective function to avoid overshoot.
When the error e(t)<0, then objective function is:

J=∫0 (𝑤1|𝑒(𝑡)| + 𝑤2𝑢2(𝑡) + 𝑤4|𝑒(𝑡)|)𝑑𝑡 + 𝑤3 𝑑𝑡 (5)

where, w4 is the weight value. The fitness function is directly related to the objective
function by the relation
1
Fitness = (6)
𝐽

4) Fitness is evaluated and fittest one is selected. The best individual reservation
strategy is adopted to improve the performance.
5) If termination criterion is not satisfied, new population N+1 is created by
reproduction, crossover and mutation.
6) The steps from 3 to 5 are repeated till l the schedule indexes are met or parameters
are convergent [7].
7

Fig.4: Flowchart of Classic Genetic Algorithm based PID tuning

5 Results and Discussion

The properties of rise time and percentage overshoot has been compared with
conventional PID controllers and FLC based PID controllers which has been given
in the following table3 [8 9 10].
8

Table 4: Parameter comparison between conventional and FLC based PID [11]

Properties PID FLC based


PID(FPID)
Rise time 30.99 7.85
Percentage Overshoot 1.28 13.66

In addition to that, the properties of Kp, Ki, Kd as well as rise time has been
compared with ZN method of tuning of PID controllers and GA based PID
controllers which has been given in the following table 5.

Table 5: Parameter comparison between conventional and GA based PID [12]

Properties ZN method of tuning GA based PID


Rise time 601.6 487.2
Kp 28.4 62
Ki 74.1 134
Kd 2.72 16

It is concluded that the adaptive PID controllers involving Fuzzy logic Controller
(FLC), Genetic Algorithm (GA) give better performance parameters while tuning
as compared to conventional ZN method.

6 Conclusion

In this paper extensive review have been carried out to list most of the machine
learning based techniques for designing of intelligent controller in applications to
autonomous vehicle. It is concluded that the adaptive intelligent PID controllers
involving Fuzzy logic Controller (FLC), Genetic Algorithm (GA) ensure improved
performance parameters while tuning the parameters of PID controller as compared
to conventional ZN method. There are lot of other optimization techniques, such as
Particle Swarm Optimization (PSO), Advanced Genetic Algorithms (AGA) that can
be effectively applied for tuning the parameters of PID controller. Intelligent
Fractional PID controller may be designed to achieve better performance

References

1. de Souza, L. B., Soares, P. P., Mendonça, M., Mourhir, A., & Papageorgiou, E.
I. (2018, July). Fuzzy Cognitive Maps and Fuzzy Logic applied in industrial
processes control. In 2018 IEEE International Conference on Fuzzy Systems
(FUZZ-IEEE) (pp. 1-8). IEEE. Rao, P. G. K., Subramanyam, M. V., &
Satyaprasad, K. (2013). Model based Tuning of PID Controller. JoCI, 16, 22.
9

2. Rao, P. G. K., Subramanyam, M. V., & Satyaprasad, K. (2013). Model based


Tuning of PID Controller. JoCI, 16, 22
3. Muhammad, Z., Yusoff, Z. M., Kasuan, N., Nordin, M. N. N., Rahiman, M. H.
F., & Taib, M. N. (2013, August). Online tuning PID using fuzzy logic controller
with self-tuning method. In 2013 IEEE 3rd International Conference on System
Engineering and Technology (pp. 94-98). IEEE.
4. Åström, K. J., & Hägglund, T. (1995). PID controllers: theory, design, and
tuning (Vol. 2). Research Triangle Park, NC: Instrument society of America.
5. Meshram, P. M., & Kanojiya, R. G. (2012, March). Tuning of PID controller
using Ziegler-Nichols method for speed control of DC motor. In IEEE-
international conference on advances in engineering, science and management
(ICAESM-2012) (pp. 117-122). IEEE
6. Kha, N. B., & Ahn, K. K. (2006, May). Position control of shape memory alloy
actuators by using self tuning fuzzy PID controller. In 2006 1ST IEEE
Conference on Industrial Electronics and Applications (pp. 1-5). IEEE.
7. de Souza, L. B., Soares, P. P., Mendonça, M., Mourhir, A., & Papageorgiou, E.
(2018, July). Fuzzy Cognitive Maps and Fuzzy Logic applied in industrial
processes control. In 2018 IEEE International Conference on Fuzzy Systems
(FUZZ-IEEE) (pp. 1-8). IEEE.
8. Kumar, T., Mishra, S. K., & Swain, S. K. (2017, March). Design of fractional
order controllers satisfying frequency domain specifications for magnetic
levitation system with time delay. In 2017 IEEE International Conference.
9. Sain, D., Swain, S. K., Saha, A., Mishra, S. K., & Chakraborty, S. (2018). Real-
Time Performance Analysis of FOI-PD Controller for Twin Rotor MIMO
System. IETE Technical Review, 1-22 de Souza,
10. Acharya, D. S., Mishra, S. K., Ranjan, P. K., Misra, S., & Pallavi, S. (2018,
November). Design of Optimally Tuned Two Degree of Freedom Fractional
Order PID Controller for Magnetic Levitation Plant. In 2018 5th IEEE Uttar
Pradesh Section International Conference on Electrical, Electronics and
Computer Engineering (UPCON) (pp. 1-6). IEEE
11. Wijesinghe, Y. M., & Narayana, M. (2019, July). Genetic Algorithm Tuned PID
controllers for Multivariable Level Control System. In 2019 Moratuwa
Engineering Research Conference (MERCon) (pp. 175-180). IEEE.
12. Butkiewicz, B. S. (2000, October). System with hybrid fuzzy-conventional PID
controller. In Smc 2000 conference proceedings. 2000 international
conference on systems, man and cybernetics.'cybernetics evolving to systems,
humans, organizations, and their complex interactions'(cat. no. 0 (Vol. 5, pp.
3705-3709)IEEE.
10

View publication stats

You might also like