In recent years, the autonomous driving industry has witnessed rapid advancements, fueled by artificial intelligence, big data, sensor technology, and high-performance computing platforms. From initial driver assistance systems to the latest end-to-end solutions, and now to high-level intelligent driving, the evolution of keywords in this sector mirrors its technological progression. As a buzzword frequently mentioned at major automakers' press conferences, 'high-level intelligent driving' signifies that the industry is accelerating towards its next phase. This term not only surpasses traditional advanced driver assistance systems (ADAS) in functionality but also showcases groundbreaking advancements in core technology, system architecture, and overall safety, guiding the future trajectory of autonomous driving technology.

However, recent regulatory developments have instructed automakers to avoid vague or exaggerated terms like 'automatic', 'autonomous', 'intelligent driving', and 'high-level intelligent driving' in promotions. This poses specific requirements for the standardized development of the intelligent driving industry. Why have these hot keywords in the industry been toned down after entering 2024? What specific intelligent driving scenarios do they refer to? Today, we delve into this concept to provide clarity. Although 'combined assisted driving' is the mandated term for promotions, based on conventional understanding, this article retains 'high-level intelligent driving' for definition.

What is High-Level Intelligent Driving?

High-level intelligent driving leverages advanced sensor fusion, intelligent algorithms, and high-performance computing to achieve a comprehensive perception, understanding, and decision-making of complex road environments, thereby reaching a higher level of autonomous driving compared to traditional ADAS systems. Beyond hardware upgrades, it undergoes deep innovations in software algorithms and system integration, enabling vehicles to navigate assisted driving in a broader range of scenarios.

When autonomous driving technology first emerged, ADAS systems primarily focused on basic driving safety functions like adaptive cruise control, lane departure warning, and automatic emergency braking. These systems enhanced driving safety but still relied on the driver's real-time monitoring and intervention. With advancements in LiDAR, millimeter-wave radar, high-definition cameras, and enhanced deep learning and big data processing capabilities, autonomous driving technology has evolved from auxiliary functions towards high intelligence. High-level intelligent driving emerged from this backdrop, aiming to enable vehicles to make decisions and execute actions in complex road environments through comprehensive sensor fusion, multimodal data processing, and advanced decision-making algorithms.

High-level intelligent driving systems require panoramic and all-weather environmental perception capabilities, involving real-time monitoring of multiple directions and precise recognition and prediction of dynamic objects. These systems often integrate LiDAR, millimeter-wave radar, high-definition cameras, and ultrasonic sensors, using multi-sensor fusion technology to mitigate single-sensor limitations in harsh weather, low light, and occlusion conditions. Deep neural networks and image processing algorithms analyze sensor data, achieving accurate detection and classification of traffic signs, vehicles, pedestrians, and other obstacles.

Differences Between ADAS and High-Level Intelligent Driving

ADAS systems primarily address basic driving assistance issues, often based on preset rules and simple sensor fusion. Their architecture is relatively simple, and functions are limited to specific driving scenarios and single tasks. In contrast, high-level intelligent driving undergoes comprehensive hardware upgrades and achieves breakthroughs in software algorithms, system redundancy, and fault self-diagnosis, capable of executing multiple ADAS functions simultaneously and adapting to diverse traffic environments. High-level intelligent driving systems must complete data collection, processing, scenario understanding, and decision-making planning within milliseconds, demanding high-performance computing platforms and real-time response capabilities. These systems typically employ dedicated AI chips and edge computing technology to ensure real-time response to emergencies during high-speed driving or in complex environments. Additionally, they incorporate redundancy designs and safety self-diagnosis mechanisms to handle hardware failures, software anomalies, or cyberattacks, ensuring a swift transition to safe mode under any circumstances.

Autonomous driving technology is classified according to the SAE standard, ranging from L0 to L5. Most current ADAS systems are at L1 or L2, requiring constant driver monitoring. High-level intelligent driving systems aim for near-L3 automation, enabling vehicles to make decisions and control themselves in limited scenarios, with the driver monitoring the environment and operating the vehicle only in specific situations. Achieving this requires high static environment perception and the ability to predict other road users' behavior through deep learning algorithms in dynamic scenarios, ensuring timely and reasonable responses. While theoretically safer and more stable, high-level intelligent driving still faces challenges in ensuring system robustness in extreme environments and handling rare events.

Applications and Challenges of High-Level Intelligent Driving

Technological breakthroughs have showcased high-level intelligent driving's performance in road tests. On highways, these systems achieve stable lane keeping, adaptive cruise control, and automatic lane changing through precise positioning, high-definition maps, and V2X information. In urban settings, multi-sensor data fusion and deep learning algorithms enable vehicles to recognize pedestrians, non-motorized vehicles, and unexpected situations, making decisions to avoid or slow down in advance. This data-driven, real-time feedback system significantly improves driving safety and paves the way for fully autonomous driving. While drivers' supervision and intervention are still required in complex scenarios, high-level intelligent driving's reliability and safety are widely recognized within the industry.

Driving scenarios include adverse weather conditions like heavy fog, rain, ice, and snow, which impact sensor data accuracy and real-time performance. High-level intelligent driving systems must safely navigate these environments. While multi-sensor fusion technology alleviates single-sensor limitations, ensuring accurate data collection and real-time processing in extreme weather or varying lighting conditions remains a challenge. Additionally, the systems demand high computing resources, necessitating low latency and real-time response while maintaining high accuracy and robustness. Edge computing, heterogeneous computing platforms, and dedicated AI chips will support future solutions, but balancing power consumption, cost, and performance requires substantial experimentation and optimization.

In terms of safety, high-level intelligent driving systems employ multiple redundancy mechanisms and self-diagnostic technologies to switch to safe mode in case of anomalies. The system typically includes multiple sets of sensors, computing units, and communication modules, ensuring basic functions even if some hardware fails. Addressing cybersecurity and data privacy, the industry is strengthening encryption technologies and protective measures. Ensuring system safety, stability, and high intelligence levels in autonomous decision-making is crucial for high-level intelligent driving technology.

Regulatory, ethical, and standardization issues also challenge high-level intelligent driving's promotion and application. Since these systems can autonomously decide in certain scenarios, issues like accident liability, ethical judgments, and data security extend beyond traditional ADAS systems. Governments and standards organizations are formulating regulations and standards, but the rapid pace of technological development outpaces regulatory progress, creating uncertainties in implementation. Ensuring societal trust in autonomous driving systems while promoting innovation and clarifying accident liability are essential for large-scale commercialization.

Long-term, high-level intelligent driving represents not just a technological upgrade but a new travel concept and intelligent transportation ecosystem. Future autonomous driving will extend beyond individual vehicle intelligence, achieving coordinated optimization through V2X communication, big data analysis, and the Internet of Vehicles. Information sharing and collaborative decision-making will smooth traffic flow, reduce accidents, and enhance energy efficiency and environmental friendliness. Core technologies like multi-sensor fusion, deep learning, and high-definition maps will continue driving autonomous driving towards full autonomy (SAE L4/L5 levels).

As a new term in autonomous driving, high-level intelligent driving signifies the industry's transformation from traditional ADAS to higher intelligence. In the quest for safe, efficient, and intelligent transportation, high-level intelligent driving is becoming the mainstream. Through road testing and data validation, the industry is refining algorithms, optimizing hardware, and accumulating practical experience. With advancements in edge computing, heterogeneous computing platforms, and dedicated AI chips, high-level intelligent driving is poised to achieve higher autonomous decision-making and intelligent control, breaking new ground in fully autonomous driving.

Summary

High-level intelligent driving not only embodies the pinnacle of current autonomous driving technology, but it also heralds the dawn of the fully autonomous driving era. At its core lies the development of a safe and efficient intelligent driving system, which leverages advanced sensor technology, multimodal data fusion, deep learning decision-making, and high-definition map positioning. Despite the current technological gap from full unmanned driving, continuous advancements, growing datasets, and collaborative efforts are progressively bridging this divide, poised to achieve a transformative leap from assisted to fully autonomous driving. It is anticipated that in the near future, intelligent transportation systems anchored in high-level intelligent driving technology will not only drastically enhance travel safety and efficiency but also fundamentally reshape travel patterns and urban traffic ecosystems, heralding a new epoch of intelligent mobility.

As the industry accelerates its development and deployment, the significance of industry standardization cannot be overstated. Promoting high-level intelligent driving must be approached with caution, and misleading marketing should be strictly prohibited. Ultimately, the widespread adoption of autonomous driving technology hinges on ensuring robust safety standards. Currently, the definition of high-level intelligent driving remains relatively ambiguous, and consumer understanding is limited. Perhaps a more apt term would be 'combined assisted driving' to better reflect its capabilities.

-- END --