From the perspective of domestic patent applications, a patent analysis is conducted on the technological evolution of the perception module in intelligent driving, based on 2 branches: sensor hardware technology and multi-sensor fusion technology. The study finds that multi-sensor combinations of cameras, LiDAR, and millimeter-wave radar remain the mainstream configuration. Feature-level and decision-level fusion technologies, due to their lower computational resource demands, have become the preferred solutions for the current large-scale deployment of intelligent driving. In the future, performance improvements in perception systems will advance along 3 directions: self-supervised spatiotemporal alignment, lightweight deployment of end-to-end models, and robustness against extreme weather conditions.

Intelligent driving  /  Perception model  /  Multi-sensor fusion