An enhanced positioning algorithm module for low-cost GNSS/MEMS integration based on matching straight lane lines in HD maps

Automated driving technology relies heavily on continuous, high-precision, and high-reliability positioning in complex urban situations. With the mass production of autonomous vehicles, the issue of keeping equipment as cheap as possible while maintaining precision is getting much attention. The low-cost global navigation satellite system (GNSS)/micro-electro-mechanical system (MEMS) integrated navigation system, as the mainstream solution, is often faced with MEMS dead-reckoning rapid divergence during GNSS outages. Meanwhile, the high-definition maps (HD maps), as the positioning resource which stores the absolute coordinate information, cover urban roads more and more perfectly. In this case, we propose a method to enhance low-cost GNSS/MEMS integrated positioning by using a monocular camera and straight lane lines in HD maps. The positioning results are corrected by projecting lanes from OpenDRIVE maps onto the image, matching them with the visually detected ones, and minimizing the reprojected linear coefficients residuals. The algorithm can be embedded into GNSS/MEMS integrated navigation system as a module without breaking the architecture and increasing the cost. The feasibility and influencing factors were analyzed in simulation and field experiments. The tests demonstrate that positions in the lateral and up directions can be corrected to within 10 cm under current HD maps accuracy. Moreover, it is not seriously limited by the initial pose error, the number of lanes, and the road flatness. The results indicate that the lane coefficients-based algorithm is a potential module for enhancing the low-cost GNSS/MEMS integrated positioning performance.