2. Key Laboratory of Opto-Electronics Information Technology of Ministry of Education, Tianjin University, Tianjin 300072, China
Overview: With the preparation of intelligent driving into industrialization and commercialization, LiDAR has become an indispensable environmental sensor with its excellent performance and has developed rapidly in hardware technology and application algorithms. This paper introduces the key technologies of intelligent driving vehicle LiDAR hardware by using LiDAR scanning method and related technology as the entry point, discussing the principle, characteristics and current status of mechanical, hybrid and all-solid-state automotive LiDAR; three points of vehicle LiDAR application algorithms, point cloud segmentation, target tracking and recognition, real-time location and map reconstruction are summarized and analyzed. Mechanical vehicle LiDAR is the earliest LiDAR product used in intelligent driving. It is still widely used in intelligent driving test vehicles because of its simple principle, easy to drive and achieve horizontal 360° scanning. The cost and unreliability of long-term use in the driving environment hinder its promotion and popularity. The hybrid vehicle-mounted LiDAR makes the mechanical structure miniaturized and electronically designed; the main components are achieved by the chip process, which is technically easy to implement and has begun to be applied in the intelligent driving solution, but its performance still needs to be optimized. The all-solid-state on-board laser radar does not have any macro or micro moving parts inside, which is reliable and durable. However, the manufacturing process is difficult and still in the early stage of development. The purpose of the LiDAR application algorithms is to accurately and reliably sense the surrounding environment and thus ensure safe and efficient driving. Among them, the point cloud segmentation algorithm is the basis of target tracking and recognition. Target tracking and recognition will realize the judgment of the motion state and geometric features of obstacles around the car. SLAM will realize the precise positioning and passable path planning of the car. The existing automotive LiDAR application algorithms have different degrees of limitations. Firstly, the accuracy and real-time application of the algorithm are difficult to satisfy at the same time. Secondly, the algorithms are mostly developed for a specific scenario, and it is difficult to ensure portability and stability. The complexity and diversity of the scene make the research of the algorithm colorful, showing a multi-level, multi-angle and thus multi-faceted situation. The analysis shows that the vehicle LiDAR will further move toward solid-state, intelligent and networked in order to reduce costs, improve performance and meet intelligent driving requirements. While, the pursuit of application algorithm research is real-time, efficient and reliable.