Introduction
The field of Artificial intelligence (AI) is one of the newest areas in science and engineering. Intelligence can be measured in terms of rationality, indicate the ideal performance, or in terms of fidelity to human understanding. When explained in terms of thinking critically, AI is the desired outcome of human effort to make computers think, portrayed as machines with minds in the literal sense. When in action, AI is the art of developing machines that execute tasks which require human creativity. When described as reasoning, AI is the study of computer processes that make it possible to comprehend, think, and act. When illustrated as acting rationally, AI is the study that integrates intelligent behavior in artifacts.
Artificial intelligence is exhibited by computational appliances and does not necessarily involve consciousness and emotions. Thus, AI is used to style computers that copy intellectual operations that people relate to a person’s mind, including studying and unraveling problems (Manoharan, 2019). Artificial intelligence has been subdivided into different fields, working independently and founded on technical concerns like particular objectives, e.g., robotics, using specific tools.
The modern abilities associated with artificial intelligence include:
- Human speech understanding.
- Completing higher levels in strategic gaming systems like chess.
- Imperfect information systems like self-driving cars and content supply networks, and military imitations.
AI techniques have gained a revival with increased computer ability and massive amounts of information in the twenty-first century. Artificial intelligence techniques have been incorporated in the technology industry, involved in solving computer science, operations research, and software engineering problems.
Autonomous vehicles have been researched and have been on the rise in recent years. They have been shown to offer many benefits to society, from improving traveling to supporting the environment’s safety. It is essential to consider that when AI vehicles can minimize human driving risks, the future will be realized if anticipatory research is conducted today (Manoharan, 2019). The paper aims at discussing the description of an AI vehicle, methodologies used in AI cars, practical analysis of the AI cars, the advantages and disadvantages, and future research about AI in self-driving cars.
Background
The aspect of autonomous cars may look to have an impossible future. Still, from the evolutionary path from Leonardo da Vinci’s invention to the current ones, there is hope that it could be achieved. Companies like Mercedes, Tesla, and Ford are competing to develop autonomous vehicles to transform the consumer world. (Batista, 2018). For instance, Ford has increased its investment by triple in its autonomous vehicle project and is conducting tests on 30 autonomous vehicles in California, Arizona, and Michigan, considered Ford Fusion hybrids.
History of Autonomous Vehicle Technology
Leonardo da Vinci was among the first people to practice Artificial intelligence about 1500 centuries before the automobile. He designed a cart that moved without exerting any force on it. Springs subjected to high tension generated power to the cart, and its steering was set in advance, and the cat moved along in an already determined path. In 1868, Robert Whitehead then invented a torpedo that was able to propel itself underwater(Batista, 2018). Later, under his guidance, it resulted in the development of other weapons and autonomous devices.
In 1933 Mechanical Mike aircraft autopilot was designed by Gyroscope Co., who used Mechanical Mike as a prototype autopilot that fled a plane for 13000 miles around the universe and tracked the plane’s movement and integrated it with controls that ensured accurate direction. Teetor, an engineer in 1945, developed a cruise control which was aggravated by the fact that he experienced a rocking motion while driving a car with his lawyer. James Adams developed a self-driving wheeled vehicle in 1961 and was fitted with cameras, programmed to move in a specific solid white line on the ground (Batista, 2018). Currently, this technology has been observed to be vital in autonomous vehicle technology.
Tsukuba Mechanical engineering Company in Japan designed an autonomous passenger vehicle in 1977 that was capable of recognizing markings on the city streets while moving at a speed of about 20 miles per hour. In 1987, Ernst Dickmanns, designed micro-processing modules to detect objects on the streets. His major invention was on imaging vital to self-driving cars to determine potential dangers and locations (Batista, 2018). General Atomics MQ-1 Predator was designed in 1995; it was a crewless aircraft that moved around the world hotspots for 14 hours. It’s integrated with technologies that have been used by cars, for instance, radars that can detect clouds, imaging cameras with thermal capabilities, and moved at dusk.
The U.S Department of Defense Research Arm (DARPA) funded several competitions that spearheaded this technology from 2004 to 2013. In 2004, a challenge competition was conducted to test navigation on approximately 150 miles of arid paths. None of the cars completed the route, but these challenges have helped discover many capabilities (Simons, 2020). The challenge of 2007 led to a prolonged urban environment where four cars could allot a six-hour time limit. In 2015, Tesla company designed a semi-autonomous Autopilot feature that has led to hands-free control and freeway driving for highways and was in the form of a single software update (Simons, 2020). Additionally, the same year, the University of Michigan’s MCity was launched as a world-class test organization for autonomous vehicle technology, and ford became the first to test their vehicles.
Levels of autonomy in self-driving cars
The U.S. NHTSA has automation levels from level 0 to level 5, and people drive through driver assistance technology of autonomous vehicles. They include Level 1, a modified driver Assistance system (ADS) that helps the driver with braking, steering, speeding but not at the same time. The vehicle entails rear-view cameras, seats that vibrate, issuing caution alerts on shifting out of the traveling lane. Level 2 is an ADS capable of steering braking or accelerating simultaneously, whereas the operator is aware but acts as the driver. Level 3 is capable of conducting operational roles in specific scenarios like car parking. The operator must be cautious about retaking charge in such situations, whereas he remains the primary vehicle driver. Level 4 can conduct all operational roles and observe the driving environment in various conditions. Under such circumstances, the human operator doesn’t need to concentrate fully. Finally, Level 5 motors ADS performs all the driving in all situations. And the people present are only passengers but not drivers.
Disadvantages of AI in Self-driving car
The AI technology used in autonomous cars has the following demerits: Self-driving cars without human drivers could frighten those experiencing it for the first time (Batista, 2019). Hacking potentials can arise that could lead to interference with this project, loss of jobs since most people depend on driving activities as their source of livelihood, loss of driving skills among most people due to the automation of the technology; hence the skills may fade with time, the technology is not suited for all-weather for instance during winter, and rainy season it could pose a great challenge and finally a lot of resources is being used in the projects and will be required due to the need of new road layouts and infrastructure.
Details and Description
AI technologies are used to power the driverless car system. Manufacturers of these cars utilize an enormous volume of information from image recognition structures and neural networks, and AI to make autonomously driverless systems. The neural networks recognize pattern data that is nursed to the AI algorithms. The information fed includes images from interior cameras in driverless cars. The neural network then uses these images to identify the surroundings like curbs, pedestrians, street signs, traffic lights, trees, and other driving environment components.
The Waymo Google autonomous car project is an example of driverless car technology. The vehicle uses a combination of devices, lidar-light recognition, and vacillating expertise comparable to sensor and camera. It syndicates data generated by the structures to identify the surroundings and forecast what those items might do next in a fraction of seconds. The more the structure runs, the more data a system can integrate into the deep learning algorithms, making more nuanced driving adoptions. The following is how the driverless car system works; the passenger sets a destination, and the vehicles calculate the distance. The lidar device examines a 60 meters array on the car, and an active 3D map is created on the current car’s environment. The sensor is fitted on the left rear wheel and examines the drive detecting the car’s location concerning the 3D maps. Distance to the obstacle is calculated by the radar system positioned on the rear and front bumpers.
All the sensors are connected with AI software, collecting input from interior video cameras and google street view. It stimulates the human perceptual, makes decisions using the control schedules, and deep learning in the driver’s regulator system such as brakes and steering. In facilitating advanced notice of things like traffic signs and lights, the system consults with google maps. The system also has an override function that enables humans to control the car (Simons, 2020). Available features in driverless cars by 2019 are hand-free steering that centers the vehicle without the driver facilitation. However, the driver must pay attention, ACC down to a stop that inevitably preserves the selectable distance between the car in front and the driver’s vehicle, and lane-centering steering that mediates when the driver crosses lane patterns by robotically bumping the car toward the opposite lane marking.
Methodology
The autonomous navigation of robot cars is attained through continuous interaction between intelligence, perception, and action. It needs the implementation and derivation of well-organized sensors that are real-time and based on controllers. Successful control algorithms for this technology should emulate the means humans are operating the human-crewed vehicles. Also, the fuzzy logic technique is applied whereby it entails reasoning that is probabilistic rather than fixed. Undefined logic variables may contain a truth value ranging from 0 to 1. It can also handle the concept of partial truth. Its truth value may vary from entirely true to completely false (Manoharan, 2019). The system can alter and affects its environment by reacting through the effectors. An appropriate rule is selected in the fuzzy controller that later provides the asymptotical system stability.
Practical Analysis of AI in Driverless cars
The main practical application of AI includes; scene recognition, lane recognition, obstacle recognition, etc. Obstacle detection technology has been used to analyze the target area, which may block the vehicle’s normal driving. The process has been facilitated by sensors such as; vision sensors, ultrasonic sensors, radar sensors, and leader sensors fitted on the driverless car. The technology is based on vision sensors and comprises monocular and binocular visions (Manoharan, 2019). A practical analysis is done to ensure that binocular vision obtains the 3D information of the scene directly and supplementation of the geometrical relationship between obstacle and road surface.
Scene classification integrates information of the whole image and their relationship to holistic output. On the other hand, scene understanding divides meaningful regions labeling them with different semantic classes. Practical analysis in lane detection is divided into two categories; one stage and two-stage methods. In the one-stage method, parameters are directly output above the lane through the deep network (Simons,2020). The two-stage process involves two steps, i.e., semantic subdivision, which is done through the deep network to output the lanes’ pixel assembly. A curve through these pixels is installed to obtain the lane parametrization.
Advantages
Improved Road Safety
Humans fear the issue of technology taking over completely and don’t want to surrender control. It must, however, be understood that self-driven autonomous vehicles improve the safety of traveling and reduce accidents. The U.S. Department of Transportation (NHTSA) reported that a more significant percentage of road accidents were attributed to human errors and poor decisions. Other causes like environmental factors and machine malfunctioning contribute to a tiny percentage of the casualties. With the introduction of self-driving cars, human drivers will be eliminated. This interprets the reduction of human-caused fatalities that make up the larger percentage of the accidents experienced.
Reduced Traffic
With humans being the leading cause of accidents, they are also the leading cause of traffic. In most cases, it only takes a careless driver to make a wrong call, like a slam on brakes or cut another driver off a busy road to cause a traffic jam which could have a rippled effect that would go for miles. When autonomous vehicles are incorporated into the system, they will run on a network that would enable communication between them and result in a well-coordinated transport system that avoids traffic jams. The vehicles would no longer be controlled by individuals but rather a plan that would enable them to work as a unit.
Convenient Parking
It is always challenging to find a parking spot, especially in a large metropolitan city where parking lots are few with many demand vehicles. With AI self-driving cars, it can be convenient because they can park themselves. Therefore, one does not have to be present for it to park, which saves time. These vehicles are also convenient because they can park closely at tight parking lots. After all, humans won’t have to worry about getting in and out of the vehicle in a parking lot.
Environmental Benefits
These vehicles help reduce carbon IV oxide emissions, which will reduce the global warming effect. The petroleum-dependent transport system contributes a significant percentage of the nation’s climate-changing emissions. With the introduction of Autonomous vehicles that use electricity, the emissions that pollute the environment will not be experienced and improve environmental quality. The Hybrid self-driving cars would use petroleum but in smaller amounts which will also reduce their ecological effects.
Disadvantages of AI in Self-driving car
Despite this technology capturing the technology and innovation sector, it has got some demerits: Self-driving cars without human drivers could frighten those experiencing it for the first time (Batista,2019). Hacking potentials can arise that could lead to interference with this project, loss of jobs since most people depend on driving activities as their source of livelihood, loss of driving skills among most people due to the automation of the technology; hence the skills may fade with time, the technology is not suited for all-weather for instance during winter, and rainy season it could pose a great challenge and finally a lot of resources is being used in the projects and will be required due to the need of new road layouts and infrastructure.
Future Research
The automotive industry is one of the most growing, especially in a digital sense. The growth has been marked by NVIDIA’s announcement in October 2017 of the world’s first AI computer to support fully driverless vehicles. The information observed the propagation of AI and resulted in an upsurge of seismic alterations to the last year’s automatic industry (Simons,2020). Researchers anticipate that 50 percent of total cars sold in 2030 with be AI driverless, which are fully autonomous. However, research on various aspects of the car has to be considered to achieve this.
Mimicking of human cognition by the car using AI
Fully driverless vehicles must have to study to cope with all the other factors. Processing data collection from numerous sources like LIDARs, GPS, ultrasonic sensors, and cameras remains the biggest challenge facing the manufacturing of fully autonomous cars. The industry’s revolving point can be to deliver car services with intuitive and cognitive abilities making sure that the new generation cars reach decisions and think like human drivers (Simons,2020). Efforts should be put in place by the manufacturers to ensure that autonomous vehicles react to abstruse conditions and consider all the possible scenarios that might influence driving. Since the aim seemed impracticable concerning location, time, and capitals, manufacturers began seeking more optimized ways of building solutions-creating an assembly convoy of cars that could learn from each other. In a fast-stride erudition situation, cars can attain the full equality of self-governance — drivers are expected able to put their eyes, minds, and hands-off on the road.
AI to Direct Uncharted Territories
To make a fully autonomous vehicle, builders still have a missing puzzle to discover. In May 2018, a significant leap was created by MIT researchers by presenting autonomous driverless cars that can navigate unmapped roads. It has a development system, MapLITE, that does not need 3D plots but depends entirely on modest GPS data collected with numerous sensors that detect the road environments. The system is beneficial, especially to communities living in rural areas.
AI to Substitute the Steering Wheels
Although many carmakers have begun their expedition to computers on wheels, few firms exist in the world at the front of making a fully autonomous experience. One of the top companies working toward this is Waymo. This Google subsidiary has since 2009 developed and today displays the uppermost level of autonomous amid the carriages handling the driving. An example of such a vehicle is their Waymo car. Though it has a steering wheel and pedals, the initial prototype presented in 2014 by google was free from one. In complying with street legal requirements, a control manual was fitted. However, google formers cars are not the only in the mentorship (Manoharan, 2019). In 2018 driverless safety report posed a zero-emanation driverless car level 4 known as Cruise A.V. It has the uppermost level of mechanization in acuity, control processes, and planning globally—this paves the way for the next-level five-car by the company in 2019.
Addition of Fuel to the Moral Dilemma by the AI
One of the complications revolving around the release of driverless autonomous cars on the street is the contentious condition that suggests life intimidating results to the individuals involved. The autonomy may result in unpredicted accidents by pushing the computers to make decisions (Simons,2020). Hence, the goal of creating a next-generation car is majorly reliant on the powerhouse of top-notch project management skills and AI-driven algorithms. The AI analyzes volumes of information from various sensors attached to the vehicle prompting decisions to run the car. It interprets the data into tangible steps supporting the next cohort of autonomous transport.
Conclusion
AI technology is used to power driverless cars by predicting the car’s surrounding environment for safe navigation. Technology has been on the rise in the recent past and promises to have a better future transport sector. The current autonomous vehicles have been an improvement of the previously discovered inventions like drones. This technology has got six levels of automation that range from level zero to five. The technology in driverless cars has impacted road safety, parking problems, and environmental pollutions. The major demerit of this technology is the loss of jobs and skills among many drivers. However, the technology is yet to be fully implemented to achieve fully autonomous cars, hence further research.
References
Batista, K. B. (2018). Self-driven cars, history, and company-driven orthodontists?The Angle Orthodontist, 88(6), 841-842.
Manoharan, S. (2019). An improved safety algorithm for artificial intelligence-enabled processors in self-driving cars. 2019, 2019(2), 95-104.
Simons, R. A., & Malkin, A. A. (2020). The forward-looking adoption timeline for driverless cars and corporate-owned fleets. Driverless Cars, Urban Parking, and Land Use, 153-173.