Lane Curve Map Evaluation for Autonomous Driving Consistency

The patent implements dynamic map updating by continuously comparing real-time map data with static map data, fitting curves to lane lines in both maps, calculating similarities, and updating the static map when consistency thresholds are met. This transforms the static map from a fixed dataset into a dynamically evolving structure that adapts to road condition changes while maintaining stability through systematic validation processes.

The patent establishes a feedback mechanism where real-time map data is continuously compared with static map data through curve fitting and similarity calculation. The system calculates fitted-curve-similarities and lane consistencies, and only updates the static map when consistency metrics meet predefined thresholds. This feedback loop ensures that real-time accuracy improvements do not compromise map consistency, as changes are validated against stability criteria before integration.

The patent segments the map evaluation process into distinct modular steps: pixel point extraction from both maps, curve fitting to lane lines, similarity calculation between fitted curves, consistency evaluation, and conditional updating. This segmentation allows the system to perform comprehensive map evaluation without overwhelming computational burden, as each module can be independently optimized and executed in a structured sequence, reducing overall system complexity.

The patent performs preliminary curve fitting to lane lines in both static and real-time maps before conducting similarity calculations. By pre-processing the lane line data into fitted curves, the system reduces the computational complexity of subsequent comparison operations. This preliminary action ensures comprehensive evaluation for reliability while minimizing processing time, as the fitted curves provide a compact representation that accelerates the consistency check process.