清华大学教授王建强：认知驱动将成智能汽车安全技术核心方向

Core Viewpoint - The development of intelligent vehicle safety technology is crucial for addressing the complex traffic scenarios and frequent accidents in China, with a focus on a cognitive-driven innovation route for high-level autonomous driving [4][5][6]. Group 1: Current Challenges in Intelligent Vehicle Safety Technology - The existing technology faces limitations due to the complexity of traffic scenarios and uncontrollable factors such as vehicle malfunctions and environmental disturbances [4]. - Current mainstream technology routes, including rule-driven and data-driven approaches, have shortcomings that hinder their effectiveness in high-level autonomous driving [4]. - Specific incidents involving Tesla, Waymo, and Uber highlight the technical shortcomings in handling unexpected and complex scenarios [4]. Group 2: Cognitive-Driven Technology as a Solution - The cognitive-driven approach is proposed as a third technological route that combines the interpretability of rule-based systems with the learning capabilities of data-driven systems [5]. - This approach emphasizes a deep understanding of the interactions between humans, vehicles, and roads, aiming to create precise models of their characteristics and operational rules [5]. - The cognitive-driven architecture consists of three layers: perception, cognition, and decision-making, enhancing reliability and adaptability in complex environments [5]. Group 3: Future Outlook for Autonomous Driving - The evolution of autonomous driving is shifting from rule-driven and data-driven methods to cognitive-driven capabilities, focusing on human-like cognition, learning, and evolution [6]. - A "three vertical, three horizontal" technical architecture is proposed to support the large-scale development of intelligent vehicles [6]. - The ultimate goal is to enhance the self-learning, self-reflective, and adaptive capabilities of autonomous driving systems, creating high-level intelligent driving systems with human-like reasoning and safety verification [6].