Autonomous driving faces various challenges such as extreme weather, dynamic obstacles and asynchronous multi-source data in special scenes. Traditional single sensor and static algorithm have problems such as low sensing accuracy and insufficient real-time performance. Aiming at the limitation of traditional method, this paper puts forward multiple sensor fusion, dynamic obstacle avoidance algorithm optimization and multimode situation awareness integration scheme. The sensor redundancy design is enhanced by the hierarchical fusion strategy, and the sensing accuracy is improved by combining the laser radar and IMU tight coupling technology; Dynamic programming algorithm and adaptive parameter modulation of reinforcement learning are introduced to realize efficient real-time obstacle avoidance; Improving image quality in adverse weather by generating countermeasure network and multi-spectral - infrared fusion technology; Integrating time-space synchronization algorithm and dynamic SLAM framework to strengthen localization robustness. This paper shows that the system significantly optimizes sensing accuracy, real-time obstacle avoidance and environmental adaptability in complex scenes, and provides a reliable technical support for the application of autonomous driving in special scenes.
Qiuyue Wang (Thu,) studied this question.