Camera-Based Lane-Aided Multi-Information Integration for Land Vehicle Navigation

Accurate positioning, especially lateral positioning, is an essential requirement of autonomous driving. Although high accuracy vehicle positioning can be conducted in open-sky environments by global navigation satellite system (GNSS), the positioning accuracy cannot be guaranteed in GNSS-denied environments such as urban canyons and tunnels. To solve this issue, inertial navigation systems, vehicle speed sensors, and vehicle motion constraints are often fused to mitigate positioning errors. However, this integration cannot meet the decimeter-level lateral positioning accuracy of autonomous vehicles' localization requirements. This article proposes a multi-information integration method aided by lane distance to further eliminate lateral error. In the proposed method, the lateral vehicle-to-lane distance measurements from camera-based systems, and the map-matching lane distance based on high-definition map were utilized to provide absolute lane distance measurement corrections in GNSS-denied environments. Field vehicular tests with low-cost integration systems were conducted to evaluate the navigation performance of the proposed multi-information integration method, and the results indicated that it is feasible for the proposed method to maintain continuous and reliable lateral positioning accuracy of better than 0.6 m under reliable lane line detection conditions in GNSS-denied environments.