1. Introduction

In the rapidly evolving world of technology, autonomous driving technology is transitioning from theoretical concepts to practical applications, bringing profound changes to the transportation sector. The development of autonomous driving technology has not been an overnight success; it has undergone a long evolutionary process. In the early days, cars relied mainly on mechanical controls, with all driving operations dependent on the driver. As electronic technology advanced, some basic driver-assistance features began to emerge, such as the Anti-lock Braking System (ABS), which enhanced safety during braking to some extent, marking the nascent stage of autonomous driving development. Subsequently, features like adaptive cruise control and lane-keeping assistance were introduced, further reducing the driver's operational burden and increasing the vehicle's level of automation. Today, highly automated driving systems are gradually making their debut, with cars capable of autonomous driving under specific conditions, signifying a major breakthrough in autonomous driving technology.

In recent years, automotive manufacturers and tech companies have significantly increased their investments in the research and development of autonomous driving technology, leading to continuous advancements in the field. From the early days of simple driver-assistance features like adaptive cruise control and lane-keeping assistance, to the current highly automated driving systems, autonomous driving technology has achieved significant breakthroughs in core areas such as perception, decision-making, and control [1]. Relevant research indicates that autonomous driving technology can not only significantly enhance road safety and reduce the incidence of traffic accidents caused by human error but also improve traffic efficiency, alleviate congestion, and provide more convenient travel options for special groups [2]. For instance, in pilot areas of some major cities, buses equipped with autonomous driving technology have seen an approximately 20% improvement in operational efficiency and a 30% reduction in accident rates, fully demonstrating the advantages of this technology.

This paper reviews the development trends and challenges of autonomous driving technology. The research covers the evolution of key technologies, such as breakthroughs in sensors, algorithms, and communication technologies, as well as practical challenges in technical, regulatory, and social acceptance aspects. This study provides a reference for future research in the field of autonomous driving technology and offers constructive suggestions for policymakers, enterprises, and researchers through an in-depth analysis of existing challenges.

2. Development Trends of Autonomous Driving Technology

With the continuous advancement of autonomous driving technology, progress in sensor technology, artificial intelligence and machine learning, as well as vehicle-to-everything (V2X) communication, has become a crucial trend in achieving higher levels of autonomy.

2.1. Sensor Technology

As autonomous driving technology progresses toward higher levels, sensor technology continues to evolve. Sensors act as the "eyes" and "ears" of autonomous vehicles, playing a vital role in ensuring their safe and efficient operation. Light Detection and Ranging (LiDAR) is seeing continuous improvements in resolution and detection range. Solid-state LiDAR, with its advantages of compact size, lower cost, and higher reliability, is expected to be widely adopted, enhancing vehicles' ability to precisely perceive their surroundings [3]. Traditional mechanical rotating LiDAR, despite its excellent performance, has limitations in large-scale adoption due to its bulky size and high cost. In contrast, solid-state LiDAR adopts a completely new technical architecture, eliminating complex mechanical rotating components, which not only significantly reduces its size but also lowers costs. Additionally, its reliability has been greatly improved, minimizing the risk of failure caused by mechanical issues. For example, a new type of solid-state LiDAR has achieved a 50% improvement in resolution and a 30% increase in detection range compared to its predecessor, enabling more precise mapping of the surrounding environment in three dimensions and providing richer, more accurate environmental information for autonomous vehicles.

At the same time, camera technology is advancing rapidly. High-resolution, wide-angle lenses combined with advanced image algorithms can more accurately identify target objects such as traffic signs, pedestrians, and vehicles [4]. High-resolution lenses can capture finer image details, while wide-angle lenses expand the field of view, reducing blind spots. Advanced image algorithms enable rapid processing and analysis of images captured by cameras, accurately identifying various target objects.

For instance, deep learning-based image recognition algorithms have achieved an accuracy rate of over 95% in recognizing traffic signs and can maintain high recognition precision even under complex lighting conditions [5]. The trend of multi-sensor fusion is becoming increasingly prominent, leveraging complementary data to provide a more robust foundation for environmental perception in autonomous driving [6]. Different types of sensors have their own strengths and weaknesses. LiDAR excels in measuring distance and precision but has relatively weaker capabilities in recognizing small objects. Cameras, on the other hand, can provide rich texture information but are significantly affected by adverse weather conditions. By fusing multiple sensors such as LiDAR, cameras, and millimeter-wave radar, the strengths of each can be fully utilized, achieving data complementarity. For example, in rainy or foggy conditions, millimeter-wave radar can compensate for the limitations of cameras and LiDAR, ensuring continuous perception of the surrounding environment by the vehicle.

2.2. Artificial Intelligence and Machine Learning

Deep learning algorithms are being increasingly applied in the field of autonomous driving. Target recognition algorithms based on Convolutional Neural Networks (CNNs) can efficiently process and analyze data collected by sensors, accurately identifying various traffic elements. By constructing multiple convolutional and pooling layers, CNNs can automatically extract feature information from images, enabling precise classification and localization of target objects such as traffic signs, pedestrians, and vehicles. For example, after training on a dataset containing millions of images, a CNN-based target recognition algorithm can achieve an accuracy rate of over 98% in pedestrian detection. Reinforcement learning allows autonomous driving systems to interact with the environment and autonomously learn optimal driving strategies, enhancing decision-making capabilities [7]. Reinforcement learning algorithms enable autonomous driving systems to learn how to make optimal decisions through continuous trial and error by setting up a reward mechanism. When the system takes actions that allow the vehicle to drive safely and efficiently, it receives positive rewards; otherwise, it receives negative rewards. Through extensive training, the autonomous driving system can gradually identify the optimal driving strategy. For example, in experiments simulating urban traffic scenarios, autonomous driving models trained with reinforcement learning algorithms reduced average travel time by 15% compared to traditional rule-based driving strategies

In the future, artificial intelligence and machine learning technologies will become more intelligent and adaptive to cope with complex traffic scenarios. As traffic environments grow increasingly complex, autonomous vehicles will need to possess stronger intelligent decision-making capabilities. Future AI and machine learning technologies will be able to perceive changes in traffic scenarios in real time and automatically adjust driving strategies based on these changes. For instance, in the event of a sudden traffic accident, an autonomous vehicle can quickly assess the surrounding environment and choose the most appropriate evasive route to avoid secondary accidents. Additionally, these technologies will have self-learning and self-optimizing capabilities, continuously improving their performance and safety as driving mileage increases.

2.3. Vehicle-to-Everything (V2X) Technology

Vehicle-to-Everything (V2X) technology is a critical enabler for the development of autonomous driving. Communication between vehicles (V2V) and between vehicles and infrastructure (V2I) will become more efficient and stable. By exchanging real-time traffic information, vehicles can plan their routes in advance, avoid collisions, and improve traffic efficiency [8]. For example, when a vehicle detects congestion ahead, it can transmit this information to following vehicles via V2V communication, allowing them to adjust their routes in advance and avoid congested areas. Communication between vehicles and infrastructure is equally important. By interacting with traffic lights, road sensors, and other infrastructure, vehicles can obtain more accurate traffic information and optimize their driving strategies. For instance, vehicles can adjust their speed based on the remaining time of traffic signals to achieve "green wave" passage, reducing waiting time at stops.

The widespread adoption of 5G networks provides high-speed, low-latency communication for vehicle-to-everything (V2X) technology, while edge computing reduces data transmission delays, enabling smarter collaborative driving for autonomous vehicles. The high speed and low latency of 5G networks allow vehicles to transmit large amounts of data, such as high-definition maps and sensor data, in an instant. Edge computing brings computational resources closer to the network edge, enabling rapid local processing and analysis of data, thereby reducing the delay caused by transmitting data to the cloud and back. For instance, in a multi-vehicle collaborative autonomous driving scenario, the use of 5G networks and edge computing can reduce information exchange delays between vehicles to under 10 milliseconds. This enables seamless coordination among vehicles, enhancing both traffic efficiency and safety.

3. Challenges Facing Autonomous Driving Technology

Despite the progress in autonomous driving technology, there are still technical bottlenecks that limit its effective application in complex and dynamic environments.

3.1. Technical Issues

Technical issues are among the most critical challenges in the implementation of autonomous driving technology. In extreme weather or special scenarios, sensors and algorithms may misjudge or fail [9]. For example, in heavy rain, the visibility of cameras can be severely affected, and the signals from LiDAR can be scattered by raindrops, significantly reducing the accuracy of environmental perception. In special scenarios, such as construction sites, the complex environment with numerous irregular objects and temporary traffic signs makes it difficult for existing sensors and algorithms to accurately identify and interpret the surroundings, leading to potential misjudgments. Current autonomous driving systems have limited computational capabilities, making it challenging to process massive amounts of real-time data quickly. The reliability and stability of these systems also need further improvement. As autonomous driving technology advances, the volume of data collected by sensors grows exponentially, placing higher demands on the processing capabilities of computing platforms. Existing computing chips often experience delays when processing data in complex scenarios, affecting the decision-making speed and accuracy of autonomous driving systems. Additionally, during prolonged operation, autonomous driving systems may encounter software failures or hardware overheating, impacting their reliability and stability.

3.2. Safety and Trust Issues

Safety is one of the core challenges facing autonomous driving technology. The public has concerns about the safety of autonomous vehicles, fearing that they may not respond correctly in unexpected situations. Establishing an effective safety evaluation system to comprehensively and objectively test and validate the safety of autonomous vehicles is key to enhancing public trust [10]. Currently, there is no unified standard for the safety assessment of autonomous vehicles, and different companies and institutions use varying testing methods and metrics. Some traditional safety testing methods, such as crash tests, primarily focus on the passive safety performance of vehicles and are not comprehensive enough for evaluating the active safety performance of autonomous driving systems. Therefore, it is necessary to establish a comprehensive safety evaluation system that covers both hardware and software, as well as normal and extreme operating conditions, to ensure the safety of autonomous vehicles is fully validated. For example, a combination of virtual simulation testing, closed-field testing, and real-road testing can be used to conduct comprehensive assessments of autonomous vehicles.

3.3. Regulatory and Ethical Challenges

Current traffic regulations are mostly based on human driving, and the emergence of autonomous vehicles has rendered many of these regulations obsolete. In terms of accident liability, it is difficult to determine the responsibilities of relevant parties [11]. When an autonomous vehicle is involved in an accident, it is difficult to determine whether the responsibility lies with the vehicle manufacturer, software developer, data provider, or other relevant parties. For example, if the accident is caused by a defect in the software algorithm, how much responsibility should the software developer bear? If the accident is due to a sensor failure, how should the vehicle manufacturer be held accountable? These issues require new regulations to clearly define responsibilities. Autonomous driving also faces ethical dilemmas, such as decision-making in unavoidable collision scenarios, for which there is currently no clear solution and requires collective discussion from all sectors of society. For instance, when an autonomous vehicle faces a sudden pedestrian and obstacle ahead and cannot avoid both, should it prioritize protecting the passengers inside the vehicle or the pedestrian? This ethical dilemma involves weighing the value of life, and there is currently no widely accepted solution.

4. Future Prospects

Despite the many challenges facing autonomous driving technology, its prospects remain promising. With technological breakthroughs, autonomous vehicles are expected to improve traffic congestion and reduce the frequency of accidents, enhancing travel efficiency and safety. In the logistics sector, the application of autonomous trucks can prevent accidents caused by human factors, reducing transportation risks and costs. In shared mobility, autonomous ride-hailing services will provide more convenient and economical travel options. By optimizing dispatching algorithms, autonomous ride-hailing vehicles can increase vehicle utilization and reduce passenger wait times. Additionally, since there is no need to pay driver salaries, operational costs will decrease, potentially lowering passenger fares.

Autonomous driving technology will also drive the development of related industries, such as intelligent transportation infrastructure construction and automotive component manufacturing, creating numerous job opportunities. To achieve these promising visions, collaboration among governments, enterprises, research institutions, and the public is essential. Strengthening interdisciplinary partnerships, improving regulations and policies, addressing technical and ethical challenges, and fostering the steady development of autonomous driving technology are all critical steps. For example, governments can increase investment in autonomous driving technology research and support related studies by research institutions. Enterprises can enhance collaboration with universities to cultivate specialized talent while the public should actively participate in discussions to share their insights on solving ethical issues.

5. Conclusion

This paper conducts research on autonomous driving technology, covering its current development status, core technological advancements and bottlenecks. It also explores future trends, including sensor technology, artificial intelligence and machine learning, and the progress of vehicle-to-everything (V2X) technology. Moreover, It analyzes challenges related to technology, safety, regulations, and ethics, and envisions future prospects. This paper promotes the application of safe and reliable autonomous driving technology, accelerates the intelligent transformation of the transportation industry, and lays a solid foundation for the development of future smart transportation systems.

However, this study has certain limitations. When discussing public acceptance, no actual surveys were conducted, and the analysis was primarily based on existing literature, resulting in a lack of in-depth and comprehensive understanding of the public's true thoughts and attitudes. Future research could collect data through questionnaires and field interviews to further explore the public's perceptions, concerns, and expectations regarding autonomous driving technology.

Looking ahead, autonomous driving technology will continue to make breakthroughs. On the technical front, sensor accuracy, algorithm performance, and computing capabilities will keep improving. In terms of applications, autonomous driving will be widely adopted in logistics and shared mobility, reshaping the transportation landscape and driving the comprehensive development of smart transportation, ultimately bringing higher benefits and value to society.