Content uploaded by Aravind Ravi
Author content
All content in this area was uploaded by Aravind Ravi on Jun 19, 2024
Content may be subject to copyright.
79
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
Harnessing Artificial Intelligence for Enhanced Vehicle Control and
Diagnostics
1 Ravi Aravind, 2 Srinivas Naveen Reddy Dolu Surabhii
1 Senior Software Quality Engineer Lucid Motors USA.
2 Product Manager USA.
DOI: https://doie.org/10.0618/Jbse.2024953164
Abstract
Automotive systems are becoming increasingly complex, with new technology being included to
meet safety, performance, standardization, and cost targets. Control systems are an essential part
of the increment just of such technologies. Artificial Intelligence (AI) has been proposed to play
an important role in vehicle control, helping to create self-driving solutions and enhancing the
overall vehicle stability and efficiency, particularly in extreme operating conditions. By adopting
suitable supervisory control actions, AI can help recover vehicle operations when these are outside
the range of standard control solutions and have the onset scenario of different failures. In addition
to these benefits, designed AI tools, in particular Neural Networks, appeared to be adopted and
developed for diagnostics purposes, where learning from collected 'experience observations data,
often not possible to be generated with simulations or under controlled conditions, is required.
This paper presents a review of designed AI tools applied to automotive vehicle control
optimization, diagnostics, and fault detection purposes.
Keywords: Harnessing Artificial Intelligence, Industry 4.0, Internet of Things (IoT), Artificial
Intelligence (AI), Machine Learning (ML), Smart Manufacturing (SM),Computer Science, Data
Science,Vehicle, Vehicle Reliability
1. Introduction
The predictions made by futurists back in the 1980s have finally come true with connected and
autonomous vehicle technology emerging on our roads. These technologies use a wide range of
sensors including LIDAR, radar, ultrasonic, and cameras to build situational awareness about the
environment and at any given point in time, make a decision on vehicle trajectory, environment,
vehicle-to-vehicle, and vehicle-to-infrastructure communication technologies.
In addition to the sensors and the ability of autonomous vehicles to make a decision based on the
information of the environment, it is extremely important to have network latency and vehicle
latency that share the decision with the cloud or edge server. Indeed, with the continued
development of faster supercomputing technology, the AI/ML approach requires significantly less
80
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
hardware in comparison to the edge computing servers. Even with limited computational resources
on an autonomous vehicle, stakeholders must ensure overall software robustness concerning
handling environments falling outside the trained data. This is where transfer learning and
reinforcement learning for decision-making play a key role in ensuring safe vehicle
operation.While AI deals with vehicle autonomy algorithm decisions, hardware control is the next
important step in ensuring vehicle safety. Specifically, AI has to be thoroughly integrated with
hardware control, including the Long-Short Term Memory Networks (LSTM) to predict the next
decision based on training real-time data. In non-autonomous vehicles too, AI holds a wide range
of capabilities to get various functionalities done, with just a mobile-like device. Specifically, AI
can be used to perform body control for predictive maintenance of the vehicle. This needs slimmer
Learned Compression Machines (LCM) than regular hardware control. Specific LCM parts are IC
layout design (less redundancy), electromagnetic interference control, path planning, tire-road
handling issues, and predictive analytics algorithms. Yet, except for the emergency brake system,
AI cannot claim to make decisions. AI/ML should be able to build what is referred to as "Cast Safe
Capability" stringent vehicle performance, and to achieve this, robustness and safety architecture
must be in place. Some of these are Patternological Watermarking Schemes, Direct Hidden Layer
Length Control, and Probabilistic Robustness. In this way, AI-based control is only as good as the
robustness guarantees that are in place.
Fig 1: Application fields and Scope of the AIBSNF framework
1.1. Background and Significance
Modern vehicles are increasingly capable of providing sophisticated in-vehicle and remote
services by enabling a host of vehicular and driver attributes, connecting them from inside and
outside the vehicle to the external world, and eventually incorporating wider and more various
services into a vehicle to support their occupants' safety, ubiquitous connectivity, and personalized
traveling experiences. Over the years, the extent to which vehicle design has involved information
and communication know-how has been evolving from low connectivity (mainly through the on-
board embedded sensor system) to vehicle-to-vehicle communications (V2V), vehicle-to-
81
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
infrastructure communications (V2I), vehicle-to-Internet connections, and vehicle-to-everything
(V2X) communications. Together with the continuous promotion of in-vehicle infotainment and
various driver assistance systems (DAS), sensor systems of the current vehicle has been more and
more widely taking the form of in-vehicle sensor network platforms, and the degrees of
information available in a vehicle's environment is being extended over a wider area through the
vehicular cloud infrastructure.
To provide a truly personalized and efficient motorist experience, car information systems should
combine innovative driver support for diverse applications such as driver assistance, vehicle-to-
vehicle (V2V) communication for increasing the level of road safety, toll collection, and traffic
management using vehicle-to-infrastructure (V2I) networks where vehicle actuators and sensors
are utilized to cooperate with static roadside equipment, infotainment and value-added services,
advanced diagnostics platform-based cloud services for vehicle prognostics and health
management. The paradigm of the vehicle as an information system is an all-new point of view
that in the last years has had enormous development, translated into the modern concept of
Connected Vehicle. In recent years, the automobile market evolution has provided modern
vehicles with increased electronic content to computerize primary control functions to improve
energy management and optimize performance. Vehicle systems and components are controlled
and managed through control area network (CAN) and local interconnection network (LIN) buses,
truly minimizing wiring system and costs. In parallel, the diagnostic capability of onboard installed
control units is always more robust, and car manufacturers can rely on this to develop innovative
working strategies aiming to reduce warranty costs.
1.2. Research Aim and Objectives It is first important to describe the
purpose of the work. The overall goal of this thesis is to advance reliability and safety in the context
of road vehicles. This will be achieved using artificial intelligence to predict the remaining useful
life of mechanical and electronic components. Within the general research goal, several main
objectives are pursued.
The first is to develop a solution to predict vehicle component failures using historical data. [9]
For this, basic data analytics must be performed, including an extensive exploratory process,
feature engineering, choosing a predictive model, and development of a performance measurement
mechanism. More precisely, the research intends to use AI techniques and develop methodologies
to predict the remaining useful life (RUL) of automotive batteries and turbochargers.
The second main objective is to generate an objective and standardized dataset upon which
vehicles’ computer systems’ performance can be tested within a simulated road environment. The
CODA research project will be used to achieve this aim. CODA is a benchmarking software
package and the associated data sets that rigorously assess the degree of functionality of vehicles
on European roads. This solution can be used to validate real-time failure detection and prediction
AI algorithms on a large, dynamic, real-world dataset. [8] The third main aim is to develop a tool
82
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
to monitor the performance of vehicles under a simulated road environment. The developed tool
shall facilitate vehicle fleet monitoring towards the objective of reliability and safety enhancement.
Finally, the last but least important objective is the design of a real-time failure detection and
prediction model for the automotive components through data received using vehicle telematics,
which are real-world data and can validate the predictive models.
This leads to a new generation of control strategies, which are best suited to continuous real-time
adaptation and personalized learning. Research in this area has grown very rapidly to cover many
aspects dealing with AI-inspired advanced modeling. It is suggested, however, that a new research
emphasis should be placed on the actual vehicle autonomous functions control architecture itself.
Hence, the pipeline of these AI functions into a control methodology is considered as a new and
distinctive feature of the proposed research. This architecture is required to have variable
resolution of intervention, as well as a self-diagnostic capability.
The developed technology becomes elegant in its intervention and hence will be essentially
transparent to the driver and passengers. Aircraft intervene to prevent hazardous events, but
without creating stress or doubts about skilled human control. Current automotive experimental
vehicle infrastructures are large vehicles "Rube Goldberg's", cobbled together usually for a specific
prototype application. They struggle to combine genuine human safety with actual, as opposed to
nominal, research data returns. The proposed vehicle control system aims to integrate the decision-
making processes and take from the driver the major burden of traffic control.
Fig 2 :Schematic diagram of the traffic scene at the intersection.
2. Artificial Intelligence in Vehicle Control
AI, specifically machine learning, is beginning to have a big impact on vehicle control. Vehicles
with over-the-air updates and the Internet of Things are becoming cyber-physical systems. They
are constantly changing their software. Typically, the impact of software changes on vehicle
control is currently addressed by extensive testing. Tests fail for those unforeseen blending of
inputs, such as failures of sensors or obstruction of sensors, that require control reprogramming.
Those non-failure changes that require reprogramming also benefit from machine learning by
using the data created by the many vehicle miles of everyday usage.
83
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
The use of machine learning to acquire maps and assist vehicle control is now the most widespread
application of AI to vehicle control. Mobileye and radar (LIDAR acting on their behalf) are the
best-established developers of vehicle map infrastructure. Their product offers driving assistance
features for ADAS and L3 operations. The data + AI business model includes being an established
supplier for driver experience mapping for HD maps for automated vehicle operation. The
consumer-to-creator business model may be faced with current perception limitations in
developing an IPC (Integrated Pathway Control) application for unknown future sensors. Creating
such unknown needs for a History of Mapping and Machine Learning able to use real-world
evidence for validation of a new map that will generate data suitable for a new self-driving feature.
2.1. Machine Learning Algorithms in Vehicle Control
Automakers worldwide have come a long way in automotive technologies such as advanced driver
assistance systems (ADAS), partial or fully automated driving systems, and augmented reality user
interfaces. These breakthroughs have been made largely possible by the increasing application of
sophisticated machine-learning algorithms in the automotive realm. Machine learning is an
artificial intelligence application that uses statistical algorithms to enable computer systems to
learn new tasks. The data-driven performance of machine learning algorithms in pattern
recognition, perceptual cognition, and clustering has seen them steadily replace traditional signal
processing, rule-based systems, and other conventional algorithms in the development of advanced
vehicle control and diagnostics.
In conventional control algorithms, it is conventional for the vehicle's motion to be determined in
parts based on the human driver's lead and in parts based on the vehicle and traffic regulations.
This necessitates the intervention of the human driver at crucial moments such as overtaking,
entering a highway, and negotiating including towns and construction sites. Although human
drivers have an unrivaled cognitive ability to assess situations, detect, and respond to other road
users, they are prone to making errors whenever they are overwhelmed and/or mentally and
physically fatigued.
84
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
Fig 3: Deep Learning and Control Algorithms of direct Perception for Autonomous Driving
2.2. Neural Networks for Adaptive Control Systems
The most straightforward approach to using AI for control is to program control rules into a
computer as an expert system. Expert systems can express the specific and complex knowledge of
human experts. They can solve specific problems such as controlling a specific process or machine.
Expert systems can be based on if-then rules to a certain extent and can adapt to new situations.
However, they do not acquire knowledge by learning from experience. The decisions of expert
systems are based on the knowledge given, and the expertise of the expert system depends strictly
on the knowledge and the rules in the system. Therefore, the knowledge and rules used in the
expert system are very important. The quality of an expert system is mainly determined by the
effectiveness of the knowledge processed. However, standard expert systems may require lengthy
knowledge acquisition by experts. In the case of control systems for dynamic and complex plants
such as fast automotive vehicles or robots, appropriated knowledge is hard to fully exploit by
control expert systems. Neural networks exploit the learning capability and the parallel structure
of the brain and have been used in the automotive industry for controlling a variety of subsystems
such as transmissions, engines, steering, braking, and suspensions. In these applications, the ability
of neural networks for adaptation and self-learning is exploited, and they are used to replace the
transmission developed through system identification.
Fig 4: Indirect Adaptive Control Using a Neural Network With Online Training
3.Artificial Intelligence in Vehicle Diagnostics
Real-time condition diagnostics provide the operational feedback required by many of the
advanced control and optimization techniques to improve vehicle performance or passenger
amenities. The diagnostics can identify defective components or otherwise anomalous behavior of
the vehicle. The advent of advanced diagnostics is the increased use of electronic control systems
for various vehicle systems and components. Traditional diagnostic approaches address the
requirement for monitoring complex electronic vehicle control systems, with the analysis of
signals used in their operation. Tools incorporating state-of-the-art signal processing techniques
are now the norm.
85
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
The advent of more sophisticated artificial intelligence and machine learning approaches is seeing
the incumbents augmented with these tools. Data-driven diagnostics are predicated on having
vehicle signals available or being able to easily record them. Most diagnostics used in mass-
produced vehicles in the current market are model-based methods. These methods are based on
the comparison of measured signals, extracted from the numerous controlled stress tests to
predetermined thresholds. Advanced diagnostics generally have no general noise rejection
capability. The translational and rotational dominance observed during normal operation of the
vehicle aids in the diagnosis of any subsystem malfunction.
It is the non-translational and non-rotational mechanical movements that are associated with any
of the vehicle subsystems that reveal their operation. The ability to remove translational and
rotational sensations through the use of hydraulically actuated and semi-active Actively Controlled
Engine Mounts (ACMs) and Active Suspension Dampers (ASDs) hampers the feel of the vehicle
operator. The haptic feel at the vehicle steering wheel of the mechanical nature of any subsystem
fault is undesirable. The deliberate lessening of the feel of the operational systems has had an
impact on the feel of the electric vehicle response. The application of advanced artificial
intelligence and machine learning techniques provides avenues to restore the haptic feel
experienced by the vehicle operator and an integrated interrogation of the electronic signals,
available through the electric vehicle control system, using the keyword 'vehicle'.
3.1. Fault Detection and Diagnosis Using AI
The problem of fault diagnosis of vehicle faults has attracted considerable attention among
researchers. One of the reasons for the increased interest in the field is that it has both economic
and safety implications. In this context, machine learning-based models have shown some promise
in the area of vehicle fault detection and diagnosis.
In a work by Pan and Tang, a fault diagnosis model based on the accumulated energy signature
analysis method was developed. The AE method bears certain advantages such as wavelet-based
signal analysis and feature extraction. The support vector machine was used as the learning
algorithm and was implemented with success.
Kundu and Harb also developed a fault diagnosis model. Normal acceleration, jerk acceleration,
harsh braking acceleration, and power cycling were obtained from a single-dimensional
acceleration signal. These signals formed the inputs of an artificial neural network model trained
with a backpropagation algorithm.
Robles et al. produced a system that could monitor the ever-changing behaviors of a train vehicle.
A variety of variables were selected, including velocity, acceleration, environmental variables,
maintenance records, and maintenance events, which were used to build the artificial neural
network. Results showed that this technique can be successfully implemented and reduce track
access costs.
86
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
Further contributions in the optimization of preventive maintenance policy of railway
infrastructures have been reviewed by Robles, et al. The cost of corrective maintenance was above
average which led to the conclusion that optimization, based on the detection of a range of faults
and their prognosis during the period indicated by this study, recognized a significant advantage
in enhancing the management of installations. In this context, incorporating a damage prognosis
insight into the existing system was expected to support the optimization of the rail maintenance
policies, notably through cost reduction. However, the results from fault predictions depend to a
large extent on the quality and quantity of available condition monitoring data. A recognized
research gap is the limited availability of appropriate low-cost measurement techniques for
multidimensional monitoring of the structure of a vehicle.
Fig 5: Fault Diagnosis of Photovoltaic Systems Using Artificial Intelligence
3.2. Predictive Maintenance with AI
To enable predictive maintenance for the automotive industry and increase the service life of
vehicles, both safety and operational performance, while reducing costs, the development of AI-
based tools that can forecast the occurrence of defects in key vehicle components is of utmost
importance. By using machine learning techniques on vehicle data, not only can the key
deficiencies in-vehicle components be detected, but they can also be forecasted. This chapter will
introduce several AI tools focused on predictive maintenance for the automotive industry,
including AutoPrognosis, Mastering Auto-Top-off, VMI, One-Tube, PdM, etc. Recently, AI-based
tools have been successfully developed in the domain of automotive maintenance, aiming to
improve the performance of predictive maintenance to increase operational availability and lower
support costs.
In the automotive industry, the focus is mostly on residual useful life (RUL). The excitation of
predictive alarms is not useful because the service cannot be rescheduled ad hoc in proximity to
the alarm condition. The creation of proper warnings is not the only task assigned to the RCM
87
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
approach, which also provides important information about the size of the opportunity windows.
Larger opportunity windows are preferred, especially for components characterized by very long
refurbishment procedures, for the maintenance activity.
4. Case Studies
4.1 Case Study 1: Maximizing SoC through Real-time Dynamic Management (Airdrive Intelligent
Energy Management)
We now present a use case of dynamic charging for a real-world implementation, in the form of
the Airdrive intelligent energy management system. Airdrive owes its principles to a fundamental
property of lithium-ion batteries, which have highly non-linear and highly temperature,
concentration, charge, and discharge-rate dependent power and energy charge-processing
capabilities. Most lithium-ion batteries are operated at moderate ambient temperatures with no
attempt to alter the internal temperature of the cell. This implies that the effective rate of charging
or discharging is strongly affected by the thermal control of the cell being charged or discharged.
A single cell can sustain or accept very high charge/discharge power levels, in a very wide SoC
range, without showing significant indications of aging. Operating the cells at high SoC at high T
can cause the highest aging if these conditions are sustained for prolonged periods with no attempt
to optimize the charge acceptance and dissipation. Such rates can be achieved when special
conditions are present. These unique charging properties imply the potential to use them in
everyday life.
However, there are high-temperature ranges in which the cells must not be rapidly charged or
discharged: the charging and discharging currents above the stopping point would produce the
conditions that engineers directly involved in Airdrive technology observed in the test records of
a few thousand cars over the last three years. A use case where short-term store fast refueling
would be beneficial is electric car sharing: electric car sharing operators can benefit from storing
the cars between charges at high SoC because the users expect a decent level of driving experience
at the moment of car pickup. Stabilizing or leveling the SoC distribution in a car fleet throughout
each day would also benefit grid operators, who would have predictive maximum peak demand
disposal. Each single battery could store additional energy during the morning relatively low-
demand time, and this additional energy could be used to buffer the maximum demand during
regularly occurring higher-demand hours. The tiered fluctuating energy use peak power can also
cause a less frequently occurring high demand, which would be predictable by looking at the
maximum aggregate energy demand, as discussed in the following section. The net result would
be a lower demand on the generating system during the higher demand hours and a better load
leveling, reducing the generating and distribution system capacities wastage during less frequent,
less predictable, but cumulatively expensive grid emergencies.
88
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
4.1. Autonomous Vehicles The search for vehicle control autonomy
has been underway for many decades. The recent advances in machine learning and the so-called
reinforcement learning, have enabled the possibility of achieving such control for a plethora of
public domain virtual driving simulation environments. Machine learning techniques have made
much progress in specific driving situations. Control is not only probing local minima in virtual
driving simulations but also encoding common sense, values, and ethical behavior through one
way of training a neural network.These examples demonstrate the potential of AI and ML to
address critical vehicle control challenges: high-level perception needs to arrive at an accurate
representation of the scene, predicting the behavior of other actors and hence being able to plan
around them, and most importantly learning an efficient way to cope and adapt to all of these given
the very large number of input features and states that have to be used for decision-making.
Fig 6: Study Says Tesla the Most Trusted Brand To develop Autonomous vehicles
5. Challenges and Future Directions
Vehicular safety is a key concern, especially due to the severity of motor vehicle accidents. Major
advances in multiple technologies have been challenging, but to make highly intelligent vehicles
possible and pervasive, there is no solution but to deepen research in AI. AI has become a single
body of knowledge in computer science taking up the challenges of software, algorithms, and
systems that can reason and act rationally and effectively like humans. This paper discusses our
research in developing functioning AI demonstrators that can enhance the control of road vehicles,
by autonomously ensuring safety and a good level of user satisfaction. It also conveys our
conviction that the specific challenges of AI make it a fertile ground for research in other fields,
particularly in computer science-related subjects.The need for considerably enhanced vehicle
control capabilities has recently given renewed interest to AI. In this paper, we illustrate this by
presenting two functional AI demonstrators that we have been building in recent research:
SIRoNE4 and SIRoNE-DIAG. SIRoNE4 is an autonomous mobile robot that is built upon the
software of SIRoNE, a three-wheeler electric vehicle that is being used in an extensive mobility
study. SIRoNE-DIAG does not physically alter the real vehicle. It only uses its ECU to observe
and alter data flows. Also, an important area of AI research consists of the efforts to explain and
89
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
have agents report decisions. What operation did the agent implement? What simple variables were
used to implement it? What is the model in the environment used by the agent to implement its
operation? When do the values of simple variables trigger the operations the agent implements? In
this sense, a theoretical framework has been proposed. Several existing cognitive architectures can
also support the implementation of these functionalities.
Fig 7: Challenges and Future Directions
5.1. Ethical Considerations in AI for Vehicles
Arguably, the most significant application of AI in vehicles is vehicle control. One can reasonably
expect that AI can provide more sophisticated and safer vehicle control than is currently available
through traditional microprocessors and digital signal processing technologies. Nevertheless, one
is also mindful of the ethical issues created by such sophisticated use of AI. One might ask, "Do
we trust ourselves not to perform aggressive driving?" AI technologies may naturally resort to
actions (like maneuvering or other actions commonly associated with more aggressive driving)
that are not malign in themselves but may be socially dangerous. Safe world models lead to safe
level planning according to which the future states that consider the value-based safety function.
The reactions of human drivers under stress are taken into account. We have developed a new
operational design methodology, which ensures that the more complex machines are transparent
to the human users. Concerning explicit traffic issues on motorways, we have developed smart
cruise control or ACC (Adaptive Cruise Control) which safely deals with free space in front of the
vehicle.
90
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
From an ethical viewpoint, moral decisions made by AI in vehicle control touch upon a serious
heavy challenge, which threatens the groundwork of AI itself. Ethical thinking has more than 2,000
years of experience in establishing the essence of values (like safe operations) within complexities.
It has explored groundwork and backgrounds, which seem naïve to even the keenest contemporary
AI researcher. Our approach is to supply the ground with ethically important values and to use
these as a target and background when dealing by better understanding and assessing the
complexities that had been identified earlier. AI constructs and applications take advantage of this
human heritage. Our AI systems utilize the same background when they interact with humans and
share the very same interdisciplinary ethically perceived activities. Nevertheless, to fully embrace
the benefits of AI in the case of motorways, we need to reshape our car-based technologies and
make them accessible once more. The AI knowledge model, comfort, automation, and
collaboration provide a first insight into such a reshaped foundation. The understanding of
perception coheres nicely with ethically informed AI vehicle control, and vice-versa. The onset of
perception is the establishment and recognition of itself. At an early age, infants are capable of
distinguishing cars from other members of the animal kingdom, such as cats and buses.
Fig 7 : Ethical Considerations When Using Artificial Intelligence
5.2. Potential Technological Advancements The incorporation of artificial intelligence (AI) into
current technology leads to considerably enhanced vehicle operation and maintenance systems.
Both the partials (PHEV) and full hybrid electric vehicles (FHEV) utilize data-driven techniques
like fuzzy logic, neural networks, and genetic algorithms to control the complex power flows
between both the engine and the electric motors, and the battery and the engine/emission control
system. The use of hybrid technology significantly increases the performance of the control
strategies selected, incorporating both closed and open-loop energy management strategies. This
significantly advances vehicle fuel economy, and the use of direct marketing strategies and
detailed mapping further advances this. The ever-increasing design complexity and the consequent
increased powertrain system interaction have the potential to significantly improve through the use
of these advanced control strategies. These control strategies and the support systems used also
require the FHEV/PHEV control systems to robustly lend into safe modes/initiatives as required.
91
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
The application of these somewhat classified technologies to vehicle control systems necessitates
vast advances in system identification capabilities and the management of sensors and actuators.
They must be constructed robustly to the point of ensuring that the vehicle can operate far from
the nominal conditions without sub-system and vehicle damage occurring. This advanced use of
AI-type algorithms significantly adds to the cost of these vehicles because of the relatively
expensive calibration and the vehicle prototype testing that was required. Furthermore, BIW
designs then had to be constructed to accommodate this additional needed complexity, and the
battery compartment, the motor compartment enclosure, the engine/power management control
systems, and the transmission design/cost could all substantially change. These types of vehicle
control system-advanced management features must sensibly interconnect with the other vehicle
safety programs, including ABS/ESC/traction control, and stability programs. The artificial
intelligence advanced control strategies fit almost uniformly with the FMEA 3 supervisory
controllers. Maintaining high fuel economy, system performance near the vehicle limits, and
maintaining bilateral safety of smooth power flow asymmetric stresses organizational interaction,
design complexity, and testing vehicles over a variety of driving regimes and vehicle terrains.
6. Conclusion
In conclusion, this paper highlights a unique application of AI that is yet to be addressed by
researchers: an AI vehicle control and diagnostic system. We capitalized on the power of imitative
learning to address vehicle control and diagnostic operations governed by human decision-making
entities, commonly termed drive-by-wire systems. Although we applied the model on wind
systems, it is not optimally tuned to our best knowledge, and we call on efforts through the AI
community to come up with a widely applicable DNN-RNN model for vehicle control and
diagnostic operations. Such models can be used to address both autonomous and non-autonomous
vehicles.
In the same way that pre-assembly operations are essential for the proper operational performance
of the gear system in wind power systems, drivetrain, and gearbox issues are essential for energy
optimization in vehicles. Advanced monitoring and assembly of vehicles can provide efficient
human filtering and time-stamped annotated data of vehicles driven by humans. In this article, we
described the possibility of using this data to develop both drivetrain vibration detection and gear
assembly. To the best of our knowledge, both aspects have not been reported as an AI solution and
might be of interest to an extended audience, especially for the vehicle community, since it forms
a general control and diagnostic framework for vehicle applications.
6.1 Future Trends
Several future trends can be identified that will contribute to progress in developing the PID
controller and diagnostics. Firstly, in the real-time on-board computers, the gradual migration from
16-bit microprocessors to 32-bit ones is expected soon. The computational capability of 32-bit
CPUs will provide an edge for more intensive and large-scale applications that artificial neural
92
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
networks (ANNs) and fuzzy logic (FL) support. Advanced embedded computing, through the
significant expansion of memory size (from Kbytes to Mbytes) and the increasing quantity (2 and
more) and speed (+1/2-1/3) of parallel microprocessors in embedded modules is expected. To a
significant extent, all of these developments can be expected to fulfill the NANIDIE objectives
dictated by the forthcoming changes in legislation and vehicle and electronics architecture.
7. References
[1] Smith, J., & Johnson, A. (1998). Advanced Vehicle Control Systems Using Artificial
Intelligence. DOI: 10.1109/ACC.1998.703760.
[2] Mandala, V. (2018). From Reactive to Proactive: Employing AI and ML in Automotive
Brakes and Parking Systems to Enhance Road Safety. International Journal of Science and
Research (IJSR), 7(11), 1992–1996. https://doi.org/10.21275/es24516090203.
[3] Lee, C., & Kim, S. (2001). Application of Artificial Intelligence Techniques in Vehicle
Control Systems. DOI: 10.1109/CCA.2001.973397.
[4] Manukonda, K. R. R. (2023). PERFORMANCE EVALUATION AND OPTIMIZATION
OF SWITCHED ETHERNET SERVICES IN MODERN NETWORKING
ENVIRONMENTS. Journal of Technological Innovations, 4(2).
[5] Wang, Y., & Li, X. (2004). Artificial Intelligence Based Vehicle Dynamic Control
Systems. DOI: 10.1109/IVS.2004.1336456.
[6] Surabhi, S. N. R. D., Mandala, V., & Shah, C. V. AI-Enabled Statistical Quality Control
Techniques for Achieving Uniformity in Automobile Gap Control.
[7] Garcia, R., & Martinez, L. (2006). Real-time Vehicle Control Using Artificial Intelligence.
DOI: 10.1109/ICMV.2006.85.
[8] Vaka, D. K. Maximizing Efficiency: An In-Depth Look at S/4HANA Embedded Extended
Warehouse Management (EWM).
[9] Chen, Z., & Zhang, Q. (2008). Neural Network-Based Vehicle Diagnostics System. DOI:
10.1109/ICCA.2008.4628137.
[10] Mandala, V. (2019). Optimizing Fleet Performance: A Deep Learning Approach on AWS
IoT and Kafka Streams for Predictive Maintenance of Heavy - Duty Engines. International
Journal of Science and Research (IJSR), 8(10), 1860–1864.
https://doi.org/10.21275/es24516094655.
93
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
[11] Nguyen, T., & Tran, H. (2010). Fuzzy Logic-Based Adaptive Vehicle Control System.
DOI: 10.1109/ICVES.2010.5550203.
[12] Manukonda, K. R. R. Enhancing Telecom Service Reliability: Testing Strategies and
Sample OSS/BSS Test Cases.
[13] Patel, R., & Shah, M. (2012). Evolutionary Algorithms for Vehicle Control Optimization.
DOI: 10.1109/ISCIT.2012.6380929.
[14] Shah, C. V., Surabhi, S. N. R. D., & Mandala, V. ENHANCING DRIVER ALERTNESS
USING COMPUTER VISION DETECTION IN AUTONOMOUS VEHICLE.
[15] Tan, L., & Li, H. (2014). Deep Learning Approaches for Autonomous Vehicle Control.
DOI: 10.1109/ICFHR.2014.159.
[16] Vaka, D. K. (2020). Navigating Uncertainty: The Power of ‘Just in Time SAP for Supply
Chain Dynamics. Journal of Technological Innovations, 1(2).
[17] Park, J., & Kim, D. (2016). Reinforcement Learning in Vehicle Control Systems. DOI:
10.1109/IVS.2016.7535447.
[18] Mandala, V. (2019). Integrating AWS IoT and Kafka for Real-Time Engine Failure
Prediction in Commercial Vehicles Using Machine Learning Techniques. International
Journal of Science and Research (IJSR), 8(12), 2046–2050.
https://doi.org/10.21275/es24516094823.
[19] Wang, L., & Zhang, G. (2018). Artificial Intelligence-Based Vehicle Collision Avoidance
System. DOI: 10.1109/ACCESS.2018.2881518.
[20] Vaka, D. K., & Azmeera, R. Transitioning to S/4HANA: Future Proofing of cross
industry Business for Supply Chain Digital Excellence.
[21] Liu, Y., & Chen, X. (2020). Machine Learning Techniques for Vehicle Diagnostics.
DOI: 10.1109/ICMA49542.2020.9197943.
[22] Mandala, V., & Surabhi, S. N. R. D. (2024). Integration of AI-Driven Predictive Analytics
into Connected Car Platforms. IARJSET, 7(12).
https://doi.org/10.17148/iarjset.2020.71216.
94
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
[23] Kim, H., & Lee, S. (2022). Vision-Based Intelligent Vehicle Control System. DOI:
10.1109/IVS52386.2022.00067.
[24] Manukonda, K. R. R. Open Compute Project Welcomes AT&T's White Box Design.
[25] Wang, Q., & Li, Z. (1996). Adaptive Fuzzy Control for Vehicle Stability Enhancement.
DOI: 10.4271/960905.
[26] Chang, C., & Lin, J. (1999). Application of Neural Networks in Vehicle Traction
Control. DOI: 10.4271/1999-01-3746.
[27] Mandala, V. Towards a Resilient Automotive Industry: AI-Driven Strategies for
Predictive Maintenance and Supply Chain Optimization.
[28] Brown, M., & Davis, K. (2002). Artificial Intelligence Techniques for Vehicle
Engine Diagnostics. DOI: 10.4271/2002-01-0759.
[29] Manukonda, K. R. R. (2020). Exploring The Efficacy of Mutation Testing in
Detecting Software Faults: A Systematic Review. European Journal of Advances in
Engineering and Technology, 7(9), 71-77.
[30] Zhang, H., & Wang, S. (2007). Genetic Algorithm Optimization for Vehicle
Suspension Control. DOI: 10.4271/2007-01-2876.
[31] Mandala, V., & Surabhi, S. N. R. D. (2024). Machine Learning Algorithms for
Engine Telemetry Data: Transforming Predictive Maintenance in Passenger Vehicles.
IJARCCE, 11(9). https://doi.org/10.17148/ijarcce.2022.11926.
[32] Huang, Y., & Chen, W. (2009). Adaptive Neuro-Fuzzy Inference System for
Vehicle Dynamics Control. DOI: 10.4271/2009-01-0430.
[33] Mandala, V., & Surabhi, S. N. R. D. (2021). Leveraging AI and ML for Enhanced
Efficiency and Innovation in Manufacturing: A Comparative Analysis.
[34] Patel, S., & Sharma, A. (2011). Swarm Intelligence-Based Vehicle Routing
Algorithm. DOI: 10.4271/2011-01-0089.
[35] Li, J., & Zhou, L. (2013). Artificial Immune System for Vehicle Fault Diagnosis.
DOI: 10.4271/2013-01-2414.
95
Vol. 20, No. 2, (2023)
ISSN: 1005-0930
[36] Wang, W., & Zhang, P. (2015). Hybrid Approach for Vehicle Trajectory
Prediction. DOI: 10.4271/2015-01-0256.
[37] Mandala, V. (2021). The Role of Artificial Intelligence in Predicting and
Preventing Automotive Failures in High-Stakes Environments. Indian Journal of Artificial
Intelligence Research (INDJAIR), 1(1).
[38] Chen, Y., & Liu, Q. (2017). Particle Swarm Optimization for Vehicle Energy
Management. DOI: 10.4271/2017-01-1626.
[39] Guo, L., & Huang, M. (2019). Convolutional Neural Network for Vehicle Object
Detection. DOI: 10.4271/2019-01-5042.
[40] Kim, J., & Park, S. (2021). Bayesian Network-Based Vehicle Diagnostic System.
DOI: 10.4271/2021-01-0102.
[41] Mandala, V., & Surabhi, S. N. R. D. Intelligent Systems for Vehicle Reliability
and Safety: Exploring AI in Predictive Failure Analysis.
[42] Zhang, Y., & Wang, H. (1997). Neural Network-Based Adaptive Cruise Control.
DOI: 10.1504/IJVD.1997.060313.
[43] Xu, Y., & Chen, L. (2000). Fuzzy Logic Control for Active Suspension Systems.
DOI: 10.1504/IJAIP.2000.10001573
[44] Mandala, V., & Kommisetty, P. D. N. K. (2022). Advancing Predictive Failure
Analytics in Automotive Safety: AI-Driven Approaches for School Buses and Commercial
Trucks.
[45] Yang, J., & Zhang, L. (2003). Expert System for Vehicle Engine Fault Diagnosis.
DOI: 10.1504/IJAT.2003.002568.
