Genetic Algorithm for Optimal Response Time Scheduling of Electric

Vehicle Model

Zouhaira Abdellaoui

and Houda Meddeb

University of Tunis El-Manar1, National Engineering School of Tunis – ENIT, Communication Systems Research

Laboratory SYSCOM - LR-99-ES21, Tunis-Belvédère, BP 1002, Tunisia

University of Lorraine, Automatic Research Center of Nancy, CRAN, UMR 7039, Cosnes and Romain, France

Keywords: DDS, Electric Vehicles, FlexRay, Genetic Algorithm (GA), Optimization, QOS, Suspension Model, SAE

Benchmark.

Abstract: Genetic Algorithms (GAs) are widely recognized for their ability to solve complex optimization problems.

Gas are an effective computational tool designed to identify optimal solutions for optimization issues in

electrical vehicle. In this context, we have developed GA for optimizing the response time based on static

scheduling suspension model of SAE Benchmark electric vehicles. The implemented architecture consists

of multiple nodes connected via the Real- Time middleware Data Distribution Service (DDS) and the

protocol FlexRay in order to benefit from their high speed and QoS.

https://orcid.org/0000-0003-2638-3911

1 INTRODUCTION

Traditional vehicles have been gradually supplanted

by electric vehicles (EVs) due to a major revolution

in the automotive sector (Putrus et al.,2009-

Clement-

Nyns and all, 2010

). This progress has led to the

development of sophisticated internal networks

within vehicles, enhancing their overall functionality

and capabilities and highlighting importance of real

time electronic systems in modern vehicle design

and operation.

In recent years, Genetic Algorithms (GAs) have

emerged as a valuable tool in the design of electric

vehicles. These algorithms are being applied to

optimize several crucial parameters. GAs have

proven to be powerful optimization tools, able to

identify solutions for scheduling problem that

enhance multiple aspects of electric vehicle

performance (Chandra et al., 2022).

In this study, we adopted GAs to optimize the

parameters that minimize the suspension model

response time in electrical vehicle. Our architecture

is based on a modern vehicle of SAE Benchmark.

This approach includes multiple nodes connected

through the the real-time communication protocol

FlexRay and the middleware Data Distribution

Service (DDS); which is elaborated in (Abdellaoui

and Hasnaoui, 2019).

Data Distribution Service (DDS) is a standard

middleware, developed by the Object Management

Group (OMG); it is used for real-time, scalable, and

high-performance data exchange between distributed

systems. DDS is frequently employed in large-scale

systems with many nodes that need to

communicate with each other, where reliability, low

latency, and efficient communication are critical

(Abdellaoui and Hasnaoui, 2019). However,

FlexRay is a high-speed communication real time

protocol used in automotive networks. It was

developed as a robust, flexibility, deterministic and

fault-tolerance communication system especially for

modern vehicles. It is very important in safety-

critical systems.

In this work, we developed a GA to estimate

optimal periods of tasks that minimize the Worst

Case Response Time (WCRT) in order optimize the

response time of suspension model of a modern

vehicle of Society of Automotive Engineers (SAE)

Benchmark; this algorithm is applied on the static

scheduling method of electric vehicles. In fact,

optimizing task response times in automotive

electrical systems can guarantee enjoyable driving

experience, performance improvement, extended

476

Abdellaoui, Z. and Meddeb, H.

Genetic Algorithm for Optimal Response Time Scheduling of Electric Vehicle Model.

DOI: 10.5220/0013276200003928

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 20th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2025), pages 476-483

ISBN: 978-989-758-742-9; ISSN: 2184-4895

battery life (Elsayed et al., 2014). The developed

architecture of GA with DDS on top of the FlexRay

bus helped us to provide a data-centric infrastructure

used in fault-tolerant system and automotive domain

whilst taking into account QoS of DDS.

This paper is structured as follows: Section 2

discusses related works. Section 3 presents the

structure design of DDS middleware QoS on

FlexRay protocol network. Section 4 is dedicated for

Genetic Algorithm method. However, the last

section discusses simulation results of suspension

model to evaluate GA performances.

2 LITERATURE REVIEW

2.1 Genetic Algorithm Approach

It is Genetic algorithms were developed by John

Holland in 1975 (

Holland, 1975).

In genetic algorithm (GAs) terminology, a

chromosome represents an individual solution. A

collection of these individuals creates a population.

Each chromosome within this population is a

potential solution for the problem to be solved

(

Chuan-Kang, 2005).The process begins by generating

an initial set of potential solutions, referred to

population. The algorithm then simulates natural

evolution by applying selection, crossover, and

mutation operators to create a new generation for

offspring solutions. The solution's effectiveness and

performance is evaluated using a problem-specific

objective function, which quantifies how well it

addresses the given challenge.

The fitness value, or objective function, of an

individual fixed its chances of survival into the next

generation. Achieving an optimal balance between

exploitation and exploration, by adjusting crossover

and mutation probabilities, can ensure high-quality

offspring and accelerates convergence in

optimization algorithms. In contrast, poorly

considered reproduction probabilities may result in

undesirable convergence to a local optimum.

This evolutionary approach allows the algorithm

to iteratively improve its solutions over multiple

generations. Genetic algorithms are effective for

many problems, including scheduling, optimization

and control.

2.2 Genetic Algorithm Applications

Genetic algorithms are considered as global search

heuristics. In fact, it is a search technique process

employed in computing to find solutions for

optimization and search problems.

GA is applied in Real Time Systems, to generate

a result which satisfies timing constraints. In

(Madureira and all, 2002), authors used GA for

assigning task priorities and offsets in order to

guarantee real time timing constraints, running on

standard Real-Time Operating System (RTOS). GAs

are also applied in planning of Robot Path based on

sensor under real-time unstructured environment

(Yasuda and Takai, 2001). Besides, job scheduling

approves again the feasibility of genetic algorithm

for the resolution of real scheduling problems, which

is solved using a set of static scheduling by GA

(Madureira et al., 2002). The authors in (Chandra

and Lalwani, 2022) implemented Genetic Algorithm

for control parameters setting optimization in hybrid

and parallel EVs. GA algorithm was proposed to

reduce FC (Engine Fuel Consumption and

emissions) using standard criteria.

3 FLEXRAY PROTOCOL AND

MIDDLEWARE DDS IN

AUTOMOTIVE NETWORKS

3.1 The Middleware DDS

We find different classes of middleware such as

DCOM, RMI, CORBA and RPC. They provide a

remote synchronous invocation method. They have

typically built on top of TCP and QoS. Also, they

are familiar with the OO programming model and

they are considered as the most-suited to closely-

coupled and smaller systems. Data Distribution

Service (DDS) is ) a real time middleware and an

open standard managed by the Object Management

Group (OMG), used as an API above operating

system (OS) and peripheral drivers that resume

common interaction patterns. DDS is the first

general-purpose standard middleware that addresses

hard real-time requirements in data-centric

applications and has a large number of configuration

parameters QoS which help developers to complete

maintainability of object state and its control in the

system. That’s why; it became actually the standard

in embedded systems. It dissociates the low-level

architecture and design of application (software

components). It isolates the design and the

validation of SW-components from hardware. It

allows the description of hardware architecture

independently of software application. Actually,

electrical vehicles have higher build complexity and

Genetic Algorithm for Optimal Response Time Scheduling of Electric Vehicle Model

477

software system than a commercial aircraft. This is

why middleware DDS is integrated, since it can

manage this complexity by reuse of software

components and exchangeability (Xiao et al, 2022).

DDS’s DCPS (Data-Centric Publish-subscribe layer)

consists of Publsiher, Subscriber, Domain-

Participant, Topic, DataWriter and DataReader

entities. The idea is to broadcast data directly from a

publisher to all its subscribers without intermediate

servers. DDS is the best middleware that can be

incorporated in an electrical vehicle to interface the

infrastructure low- level. In fact DDS have (Putrus et

al.,2009):

•

Self-healing communication: In case the

network is repaired, the network will

quickly find new nodes and will work again

thanks to the built-in discovering entities

•

Support for custom fault-tolerance:

Implementations can add freely further fault

tolerance such as FFT (Xiao et al, 2024)

•

The support of many network interfaces like

in FlexRay channels (channel-A and

Channel-B) as well as redundant Data

Readers and Data Writers on each node

results in networks completely separate.

Even in the event of a complete network

failure, the system will continue to operate.

•

No single point of failure: DDS needs

«special" nodes, there for it can be

implemented without a single point of

failure thanks to publisher and subscriber

redundancy.

3.2 Flexray Protocol

We have chosen to work within a platform of a

vehicular network based on the extended SAE

BENCHMARK. In this system; a set of network

processors subsystems produces routing data. This

data must be distributed along the vehicular

network.

In fact, we have applied the studied approaches

on a new vehicle benchmark developed in

(Abdellaoui and Hasnaoui, 2019) and based on the

SAE Benchmark. However, this Benchmark was

designed to the best fit the CAN network and with

major modifications and adjustments it be adapted to

the FlexRay protocol.

The resulting architecture is composed of 15

nodes connected by the FlexRay bus. The main

objective of this architecture that we mentioned

previously is to guarantee better performance of the

vehicular network and ensure safety and reliability.

In each node we find an embedded a Real-Time

Operatng System and a publish/subscribe

middleware; in our case we are adapting the DDS as

middleware for developed and modern vehicle. That

will be the best manner to validate of our vehicle

system design. FlexRay Networks is one of the

newest X-by Wire communication systems that offer

high speed which reaches 20 bits/s thanks to its two

channels.

4 SUSPENSION MODEL AS DDS

ENTITIES AND STATIC

SCHEDULING MODEL

4.1 DDS Entities: Suspension Model

In an electric vehicle (EV), the suspension system

plays a crucial role in ensuring ride comfort,

handling stability, and maximizing the efficiency of

the vehicle’s electric power train. Given the unique

characteristics of EVs, their suspension systems

often incorporate specialized designs to address

challenges like battery weight, energy efficiency,

and precise handling. To evaluate this methodology,

we integrated DDS middleware into the vehicle

Blockset. This integration enables applications to

communicate by publishing data they produce and

subscribing to data types they consume. The DDS

middleware allows components in the vehicle

system to exchange information in a publish-

subscribe model, facilitating efficient data

distribution between different nodes of suspension

unit.

4.2 Static Scheduling Model

The Real-time networks protocol FlexRay use either

dynamic or static scheduling approach for

management task. In static scheduling,TDMA Time

Division Multiple Access (TDMA) technique is

employed (

ISO, 199). and tasks are scheduled based

on a time-triggered approach. Each message is

allocated a fixed interval time, "slot," within a

recurring time frame. While this approach ensures

deterministic message transfer, it does not fully

exploit the available network bandwidth. In the

FlexRay protocol, all frames are designed with a

consistent and unchanging length. This

standardization ensures predictability and efficiency

in data transmission where each node employs a

sophisticated tracking system and maintains two

separate slot counters, known as vSlotCounters for

ENASE 2025 - 20th International Conference on Evaluation of Novel Approaches to Software Engineering

478

Figure 1: Application scenario: Suspension Model.

channel A and B, the two slot counters initialize at 1

and increment after each slot.Tasks in the static

segment are periodic and maintain the equal priority

across each communication cycle. The Worst Case

Response Time (WRST) calculated based on the

static scheduling model is shown below:

𝑅



=𝐶𝑆



+𝐶



+𝐵



+









𝐶𝑇



∑











𝐶𝑇





(1)

Table 1: Suspension Parameters.

𝐶𝑆

Costs associated to task switching

𝐶

𝑖

Worst-Case

computation time for tasks i,

Worst case of blocking time for the task

𝑇

𝑐𝑙𝑘

Clock period for a certain frequency

Clock interrupt cost

Periodic tasks set

𝑇

Minimum time among jobs, task

releases or task

eriods

Moving cost for single task from

dela

queue to the run queue,

5 GA FOR SUSPENSION MODEL

OPTIMAL RESPONSE TIME IN

EVS

In this work, genetic algorithms are employed to

suspension model of electric vehicles in order to

optimize the response time related to each task.

5.1 Genetic Algorithm’s Architecture

The main purpose of this research is to find

tasks period of Suspension model to minimize their

response time. Let T be a decision variable

represented as a column vector with size n, which

specifies the periods for n tasks, and let R denote the

response time as defined in equation (1). There

for Suspension model can be outlined as shown in

the given equations:

min 𝑓(𝑇) =

∑

𝑅







(T) (2)

Under constraints of

5 ≤ 𝑇𝑖 ≤ 320 (3a)

𝑅𝑖 ≥ 𝑀𝑖𝑛_𝑠𝑒𝑝 (3b)

𝑅𝑖 ≤ 𝐷𝑖 ≤ 𝑇𝑖 (3c)

In Equation (2),

(

𝑇

)

represents the fitness of each

chromosome T in the population, calculated as

the total response time across all tasks. The

objective is to identify a vector T* which minimizes

this objective function

(

)

. T defined as solution

space is further constrained by a set of conditions

outlined in Equation (3). Four operators were

taken into account for the development

the

proposed

genetic

algorithm

as shown in figure 2.

The descriptions of these functions are:

Selection: It's a mechanism designed to enhance the

probability that superior solutions will reproduce

and transmit their beneficial traits to the next

generation. This process enables the identification of

the fittest individuals within the current population

to contribute to the formation of the subsequent

generation (

Kinnear, 1994).

Genetic Algorithm for Optimal Response Time Scheduling of Electric Vehicle Model

479

Crossover: Crossover in genetic algorithms (GAs)

is a key genetic operator that combines the genetic

information of two parent solutions to produce one

or more offspring solutions. This mimics the

biological reproduction process and helps in

exploring the solution space more effectively.

Mutation: It is applied at the gene level to the

chromosomes produced by the crossover operation.

It involves selecting a gene as a mutation point and

changing its value to a random integer in the range

of [5, 320]. Here, the lower bound indicates the

minimum value for the gene, while the upper one

signifies its maximum value.

Figure 2: Genetic Algorithm Flowchart.

Rectification: An additional adjustment mechanism

was required when the genetic values of an offspring

do not meet last constraints outlined in equations (3).

This led to the creation of a rectification function

that produces a modified population, denoted as 𝑸˜.

of size M which represents the set of n tasks periods.

𝑇





∈𝑄



is the rectified individual in these conditions:

If 𝑇



<5 then 𝑇





=5, (4)

If 𝑇



> 320 then 𝑇





= 320, (5)

If 𝑇



<𝐷



then 𝑇





=𝐷



, (6)

𝑇





= 𝑇



−𝑚𝑜𝑑

(

𝑇



)

(7)

With , 𝑖 = 1,..,𝑛; 𝑘=1,….,𝑀 and the mod

is the operator which returns the remainder after

division of 𝑇



by 5.

5.2 Genetic Algorithm Application

Scenario

The scenario we have described, outlines a distributed

system for a vehicle's active suspension control

including Rear_Right_wheel, Front_Left_wheel,

Rear_Left_wheel and Front_Right_wheel,using real

time Data Distribution Service (DDS) middleware

and the protocol FlexRay networks.

The inputs for genetic algorithm employed for our

application are:



Task parameters that involve ID, data field

length, delay, deadline, clock period,

interrupt handling cost and clock interrupt.



Network parameters defined by the speed,

which vaires between 5 Mbit/s, 10 Mbit/s,

or 20 Mbit/s.

Its goal is to determine optimal task periods that

minimize response times, there by facilitating low-

latency signal transmission.

Table 2: Results of Suspension model with bus speed of 5 Mbit/s.

Tasks Message ID Size (byte) T (ms) D (ms) Min_Sep

(ms)

R (ms)

Front-leftwheel

module

1 20 5 0.0206 0.0206

23 2 10 10 0.0226 0.0227

Front-right wheel

module

10 1 20 5 0.0206 0.0206

24 2 10 10 0.0226 0.0227

Rear-leftwheel

module

11 1 75 5 0.0206 0.0206

25 2 10 10 0.0226 0.0227

Rear-right wheel

module

12 1 15 5 0.0206 0.0206

26 2 10 10 0.0226 0.0227

Active suspension unit 27 2 10 10 0.0226 0.0227

ENASE 2025 - 20th International Conference on Evaluation of Novel Approaches to Software Engineering

480

Table 3: Results of Suspension model with bus speed of 10 Mbit/s.

Tasks Message ID Size(byte) T(ms) Delay(ms) Min

separation

(ms)

R(ms)

Front-leftwheel

module

1 140 5 0.0103 0.0105

23 2 125 10 0.0113 0.0115

Front-right wheel

module

10 1 20 5 0.0103 0.0105

24 2 20 10 0.0113 0.0115

Rear-leftwheel module 11 1 140 5 0.0103 0.0105

25 2 130 10 0.0113 0.0115

Rear-right wheel

module

12 1 20 5 0.0103 0.0105

26 2 20 10 0.0113 0.0115

Active suspension unit 27 2 10 10 0.0113 0.0115

Table 4: Results of Suspension model with bus speed of 20 Mbit/s.

Tasks Message ID Size(byte) T(ms) D(ms) Min_Sep

(ms)

R(ms)

Front-leftwheel

module

1 75 5 0.0052 0.0053

23 2 45 10 0.0057 0.0058

Front-right wheel

module

10 1 160 5 0.0052 0.0053

24 2 270 10 0.0057 0.0058

Rear-leftwheel module 11 1 5 5 0.0052 0.0053

25 2 45 10 0.0057 0.0058

Rear-right wheel

module

12 1 235 5 0.0052 0.0053

26 2 105 10 0.0057 0.0058

Active suspension unit 27 2 10 10 0.0057 0.0058

6 EVALUATION OF SUSPENSION

MODEL SIMULATION

RESULTS

For this evaluation, our application is presented by

the standard SAE Benchmark model using

middleware DDS, FlexRay real-time protocol as

communication support and GA to estimate the

worst-case response time (WCRT) for

communication tasks across suspension's nodes.

6.1 Suspension Simulation Results

The genetic algorithm was performed with a

population size of 30 and 300 iterations. The results

are displayed in the following tables.

The results confirm that DDS Quality of Service

(QoS) requirements are effectively met, especially

those concerning deadline and minimum separation,

while accounting for latencies introduced by the

FlexRay. The objective function evolution for the

suspension node is illustrated in figure 3.

6.2 GA Performance Evaluation with

Suspension Model

To evaluate the GA's performance, tests were

conducted with different population sizes and bus

speeds. The table shows the results for the

suspension-model.

Table 5 reveals that convergence time tends to

increases as the population size grows. Moreover,

the proposed GA occasionally converges to a local

optimum. However, a population size of 30 yielded

the global optimum. Cost or objective function

evolution for suspension model is illustrated in

figure Figure 4.

Genetic Algorithm for Optimal Response Time Scheduling of Electric Vehicle Model

481

Figure 3: Objective function Evolution for Suspension model.

Table 5: Algorithm Performance Results in function of bus speed and population size.

Bus speed 5Mb/s et iteration 300

Population 10 30 50 70 100 120

Convergence time 10.31 59.99 80.93 168.15 312.43 451.31

Objective function 0.1962 0.1959 0.1966 0.1963 0.1961 0.1961

Bus speed 10Mb/s

Population 10 30 50 70 100 120

Convergence time 8.71 52.59 123.81 220.65 325.74 470.29

Objective function 0.0995 0.0995 0.1004 0.1004 0.1004 0.1000

Bus speed 20Mb/s

Population 10 30 50 70 100 120

Convergence time 10.62 37.06 67.50 136.54 237.65 375.69

Objective function 0.0511 0.0502 0.0502 0.0502 0.0511 0.0511

Figure 4: Objective function evolution for Suspension

model.

In a second scenario, we defined the stopping

criterion as the verification of constraint (3a). The

algorithm halts once it finds a solution that satisfies

this constraint. The adopted algorithm was

implemented for suspension nodes, and the results

for a bus speed of 5 Mb/s are presented in table 6:

Table 6: Performance of GA in function of population

size.

Bus speed 5Mb/s

Population 10 30 50 70 100

Convergence

Time

3.43 10.12 24.07 67.41 99.10

Objective

function

0.196

Iteration

number

123 90 100 127 114

ENASE 2025 - 20th International Conference on Evaluation of Novel Approaches to Software Engineering

482

7 CONCLUSION

This work introduces a novel approach of genetic

algorithm based on the static scheduling model of

SAE Benchmark electric vehicle EVs. The design

integrates the Real-Time protocol FlexRay with

Data Distribution Service middleware.

GA approach is applied on the Suspension model

to identify the optimal task periods which led to

minimize the required response time.

A performance evaluation is conducted to

validate the efficiency of the proposed approach in

the design of EVs.

ACKNOWLEDGMENTS

We wish to express our gratitude to the SYSCOM

ENIT members for their help and assistance.

REFERENCES

Putrus G. A., Suwanapingkarl P., Johnston D., Bentley E.

C., and Narayana M. (2009). “Impact of Electric

Vehicles on Power Distribution Networks”, in Proc.

IEEE Vehicle Power and Propulsion Conf., pp. 827-

831.

Clement-Nyns K., Haesen E., and Driesen J. (2010). “The

Impact of Charging Plug-In Hybrid Electric Vehicles

on a Residential Distribution Grid”, IEEE Trans.

Power Systems, vol. 25, pp. 371-380.

Chandra A., Lalwani J. and Jajodia B. (2022). “Towards

an Optimal Hybrid Algorithm for EV Charging

Stations Placement using Quantum Annealing and

Genetic Algorithms,”International Conference on

Trends in Quantum Computing and Emerging

Business Technologies (TQCEBT).

Abdellaoui Z. and Hasnaoui S. (2019). “DDS Middleware

On top of FlexRay Networks: Simulink Blockset

Implementation of Electrical Vehicle and its

Adaptation to DDS concept using the FlexRay

Protocol’’, International Journal of Soft Computing,

Soft Computing Volume 23, pages11539–11556.

Elsayed S. M., Sarker R. A., and Essam D. L. (2014), “A

new genetic algorithm for solving optimization

problems,” Engineering Applications of Artificial

Intelligence, vol. 27, pp. 57–69.

Holland J. H. (1975). Adaptation in natural and artificial

systems. Ann Arbor: University of Michigan Press.

Ting, Chuan-Kang (2005). “On the Mean Convergence

Time of Multi-parent Genetic Algorithms without

Selection”. Advances in Artificial Life: 403–412. ISBN

978-3-540-28848-0.

Madureira A., Ramos C., do Carmo Silva S. (2002). “A

Coordination Mechanism for Real World Scheduling

Problems using Genetic algorithms”, Evolutionary

Computation, 2002. CEC ’02. Proceedings of the

2002, Congress on, 1, pp 175 –180.

Yasuda, G. and Takai, H. (2001). “Sensor-based Path

Planning and Intelligent Steering Control of

Nonholonomic Mobile Robots”, Industrial Electronics

Society, 2001. IECON ’01. The 27th Annual

Conference of the IEEE, 1,Page(s): 317 -322,1.

Chandra A., Lalwani J. and Jajodia B. (2022), “Towards

an Optimal Hybrid Algorithm for EV Charging

Stations Placement using Quantum Annealing and

Genetic Algorithms,”International Conference on

Trends in Quantum Computing and Emerging

Business Technologies (TQCEBT)

Xiao H., et al. (2022)."Edge Intelligence: A

Computational Task Offloading Scheme for

Dependent IoT Application," in IEEE Transactions on

Wireless Communications, vol. 21, no. 9, pp. 7222-

7237, doi: 10.1109/TWC.2022.3156905.

Xiao H., et al. (2024). “VAAC-IM: Viewing Area

Adaptive Control in Immersive Media Transmission”.

In Proceedings of the 34th edition of the Workshop on

Network and Operating System Support for Digital

Audio and Video (NOSSDAV '24). Association for

Computing Machinery, New York, NY, USA, 8–14.

https://doi.org/10.1145/3651863.3651877.

International Standard Organization (1994), ISO 11519-2,

Road Vehicles - Low Speed serial data

communication- Part 2: Low Speed Controller Area

Network, ISO.

Kinnear K. E. (1994).’’A Perspective on the Work’’in this

Book. In K. E. Kinnear (Ed.), Advances in Genetic

Programming (pp. 3-17). Cambridge: MIT Press.

Genetic Algorithm for Optimal Response Time Scheduling of Electric Vehicle Model

483