A Two-Stage Crowdsourced Map Matching and Alignment Method Combining a 2D HD Map with Decoupled Pose-Estimation Degrees of Freedom [Patent]

This invention proposes a two-stage crowdsourced map matching and alignment method that combines a 2D HD map with decoupled pose-estimation degrees of freedom. Based on nonlinear optimization, it maximally eliminates the global-position inconsistencies among local maps uploaded by different vehicles at different times. A two-stage optimization method with decoupled pose degrees of freedom is proposed, which can effectively mine the information contained in the 2D HD map. Meanwhile, the pose degrees of freedom are selectively decoupled and estimated according to different scenarios; because the dimensionality of the estimated state is reduced, the pose-estimation process converges more easily and yields more accurate results. This produces a more tightly converged map form, reduces the difficulty of the subsequent map-fusion stage, and enables computation of a cloud-side crowdsourced fused map under more globally consistent poses.

Download paper here