Hierarchical Perception Enhancement for Different Levels of Autonomous Driving: A Review

High precision, strong reliability, and wide range of perception can provide essential data support for the safe and efficient driving of autonomous vehicles. Based on the technology and method of autonomous driving perception, and taking into account the background and current research, a three-hierarchy architecture is developed to achieve perception enhancement: autonomous perception including self-localization and visual perception, fusion perception, and cooperative perception. Depending on the perception needs and other features of different levels of autonomous driving, different enhancement methods can be applied. First, perception is enhanced by algorithm optimization mainly for driving assistance. Next, perception is enhanced by multimodal and heterogeneous data fusion for conditional autonomous driving. Then, perception is enhanced by connected and interactive data cooperation for high-level autonomous driving. If the level of autonomous driving changes, the main and primary modes of perception enhancement can also be switched. The development revealed that cooperative perception enhancement of swarm intelligence based on vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and other connected technologies is an inevitable trend. Future studies will focus on processing multisource and heterogeneous sensor data. The core is to break through the technical bottleneck of fusion and cooperation, and improve the perception ability and adaptability in complex traffic environments. Perception enhancement will accelerate the development and practical implementation of autonomous driving and ultimately improve traffic safety and efficiency.